THE ACOUSTIC AND PERCEPTUAL CORRELATES OF EMPHASIS IN
URBAN JORDANIAN ARABIC
BY
Mohammad Al-Masri
M.A. Yarmouk University, Jordan, 1998
B.A. Yarmouk University, Jordan 1994
Submitted to the graduate degree program in Linguistics and the Graduate Faculty of
the University of Kansas in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
_______________________________
Chairperson
Committee members
_______________________________
_______________________________
_______________________________
_______________________________
Date defended: _______________________________
The Dissertation Committee for Mohammad Al-Masri certifies that this is the
approved version of the following dissertation:
THE ACOUSTIC AND PERCEPTUAL CORRELATES OF EMPHASIS IN
URBAN JORDANIAN ARABIC
_______________________________
Chairperson
Committee members:
_______________________________
_______________________________
_______________________________
_______________________________
Date approved: _______________________________
ii
Abstract
Acoustic and perceptual correlates of emphasis, a secondary articulation in the
posterior vocal tract, in Urban Jordanian Arabic were studied. CVC monosyllables
and CV.CVC bisyllables with emphatic and plain target consonants in word-initial,
word-medial and word-final positions were examined. Spectral measurements on the
target vowels at vowel onset, midpoint and offset revealed a significant increase in F1
and F3 and a decrease in F2 in emphatic compared to plain environments. Emphasis
was observed more in the environment of high and low front vowels than high back
vowels, anticipatory emphasis spread was more salient than perseveratory emphasis
spread, and males showed more emphasis than females. Spectral moment measures of
the target consonants themselves revealed inconsistent or no effects of emphasis.
Results from a perception experiment with cross-spliced monosyllables showed that
perception of emphasis follows the vowel, not the consonant, in the emphatic
syllable, thus confirming the results of the acoustic experiments.
iii
Acknowledgements
Pursing a PhD degree in Linguistics would have been impossible without the
guidance and assistance of my supervisor, Prof. Allard Jongman. I would like to
express my sincere gratitude and appreciation for his help during this long, yet
enriching journey. His continued support, everlasting patience, scholarly spirit and
the many revisions of this dissertation made it possible for this humble work to come
to life. I would also like to thank many other great people: Prof. Sara Rosen for her
kind and supportive spirit, Prof. Joan Sereno for the invaluable remarks on parts of
this dissertation, and Prof. Jie Zhang for his helpful remarks before the dissertation
defense. I am also grateful for all the wonderful faculty members in the Linguistics
Department, especially Prof. Fiona McLaughlin for her faith in me. Special thanks go
to Alex Hornbrook and Corinna Johnson for their dedication and support.
I would also like to thank all my colleagues at KUPPL: Sanae Akaba, Sonja
Combest, Wendy Herd, Yuwen Lai, Kazumi Maniwa, Ariane Rhone, Eva Rodriguez,
Travis Wade and Wael Zuraiq, Many thanks also go to my friends: Osama
Abdulghafer, Khalid Abu Abbas, Abdullah Jaradat, Maria del Carmen Parafita Couto
and Sabri Shboul for the great time and discussions we had.
I would like to extend my appreciation to the great people in the Department
of African and African-American Studies, especially Prof. Naima Boussofara for the
great opportunity she gave me and for being a great mentor, and Prof. Peter
Ukpokodu and Lisa Hall for their continued support.
iv
My love and appreciation go to my great wife, Leena for her unequivocal
support and kind spirit during this experience. And to my little son, Saleh, for being
kind and nice. I am also grateful for the prayers of my wonderful family, especially
my parents.
This research was partially supported by a National Science Foundation grant
(# 0518969). Many thanks for their support.
v
Table of Contents
Abstract ………………………………………………………………………………iii
Acknowledgements …………………………………………………………………..iv
Table of Contents ………………………………………………………………...…..vi
Chapter One: Introduction and literature review ………………………..……………1
1.1.Introduction .…………….……………..……………….…………………1
1.2.Emphasis .………………………………..….…….………………………2
1.3.Urban Jordanian Arabic (UJA) ………………………..……...…….…….6
1.4. The articulatory and acoustic correlates of emphasis ………...………….8
1.5. The perceptual correlates of emphasis ………….……………...……….20
1.6. Statement of the problem …………………….……………...………….21
1.7. Rationale of the study …………………….…………..……………….. 22
1.8. Objectives of the study ……………….……………………...………….23
Chapter Two: Acoustic study – the monosyllables .…………………………………24
2.1. Method ………………………………..…………..……………….……26
2.1.1. Subjects………………………………………………………..26
2.1.2. Materials and Procedure……………………………...………26
2.1.3. Measurements………………………………..………………..27
2.2. Analysis ………………………………………………………………….29
2.3. Results ………………………….………………………………………..30
2.3.1. Duration………………………………...……………………..31
2.3.2. Formant frequency…………………………………………….33
2.3.2.1. F1……………………………...……………………33
2.3.2.2. F2………………………………………..………….39
2.3.2.3. F3 ……………………………………….………….49
2.3.3. Spectral moments………………..……………………………..55
2.3.3.1. Target consonants ………………………………….55
2.3.3.2. /b/ consonant …………………...…………………..56
2.3.4. Locus equations ………………………..………………………57
2.4. Results Summary ………………………………..……………………….61
2.5. Discussion ……………………………………...……………………..….62
Chapter Three: Acoustic Study – the bisyllables….………..…………………..……68
3. 1. Method …………………………………………………………………..69
3.1.1. Subjects………………………………….…...………………….69
3.1.2. Materials and Procedure…..…….….……………...…………...70
3.1.3. Measurements ……...……………………………………………71
3.2. Analysis ……………………………………………………..……………74
3.3. Results …………………………….……………………...……………….75
3.3.1. Duration …………………………….……………………..……..75
3.3.2. Formant Frequency ………………………..……………....……..76
3.3.2.1. F1……………………………………………………77
3.3.2.1.1. target consonant word-initial …….…….…78
3.3.2.1.2. target consonant word-medial ……………82
vi
3.3.2.1.3. target consonant word-final ………………90
3.3.2.2. F2 …………………………………………………...93
3.3.2.2.1 target consonant word-initial ……………...93
3.3.2.2.2. target consonant word-medial…………..…97
3.3.2.2.3. target consonant word-final……………...103
3.3.2.3. F3 ……………………………………………….…105
3.3.2.3.1 target consonant word-initial …………….105
3.3.2.3.2. target consonant word-medial…………....107
3.3.2.3.3. target consonant word-final……………...109
3.4. Results Summary …………………………………………..………...…..111
3.5. Discussion…………………………………………………….…………..113
Chapter Four: The perception study ……………………………….………………117
4. 1. Method …………………………………………..……………..………120
4.1.1. Subjects…………...…………………..………….…………….120
4.1.2. Materials and Procedure………..................……………..…… 120
4.2. Analysis ……………….………………….……………………….……..123
4.3. Results …...………………………………...…………………………….124
4.3.1. Original words……………...….…………...………………. 124
4.3.1.1. Summary ……..………………..……...…………….125
4.3.2. Changing target consonant …..…………..…………...……..126
4.3.2.1. Summary ………...………………………………….131
4.3.3. Changing target vowel……………..........……………………….132
4.3.3.1. Summary ………………………..…….…………… 136
4.4. Discussion ……………………….…………………….…………………137
Chapter Five: Summary, discussion and conclusion ………..…………..…………143
5.1. Summary of results……………………………………...………………..144
5.1.1. Acoustic study: the monosyllables..…….……….………………..144
5.1.2. Acoustic study: the bisyllables…………………..………………..145
5.1.3. The perception study…………...…………………….……………146
5.2. Discussion…………………………………..………………...…………..146
5.3. Conclusion ..…… ………………...………………..…………………….149
5.4. Follow-up … ..…………………………………………..………………..150
References ………….…………...………………...……………..…………………152
Appendices...………………………………………………………………………..157
vii
Chapter One
Introduction and literature review
1.1. Introduction
Emphasis is a feature of Semitic languages under which some consonants are produced
with a secondary articulation in the back part of the oral cavity. Emphasis has been
studied extensively in many dialects of Arabic, mostly through the contrast between
plain and emphatic segments. It has been dealt with as a phenomenon pertaining to
pharyngealization and/or uvularization. Both pharyngeals and uvulars are produced in
the back part of the oral cavity with slightly varying places of articulation. Many
sounds in Semitic languages are primarily produced in this region, and secondary
articulation involving this region is also characteristic of Semitic languages.
Research on emphasis dates back to the era of classical Arab linguists in the eighth
century AD. They described emphasis in terms of production with scattered attempts to
describe the vocal tract configuration during the production of emphatics. A large
number of the studies that investigated emphatics were either articulatory, focusing on
describing the configurations of the articulators involved in the production of
emphatics, or phonological, concentrating on aspects such as the feature RTR
(Retracted Tongue Root), emphasis domain, emphasis spread, and emphasis blockers.
Some of the more recent studies on emphasis focused on sociolinguistic aspects of
emphasis pertaining in particular to gender distinctions in the production of emphasis.
1
A considerably increasing number of studies have been examining the acoustic
correlates of emphasis. However, thorough investigations of the perceptual mapping of
these correlates have been very scarce. Despite the substantial literature on emphasis,
many aspects of the phenomenon are still murky due to a number of reasons such as
variations from one language/dialect to another, the design of the experiments
conducted, the number and dialectal background of the subjects that participate in these
studies, limitations on the consonants and vowels used in the data, to mention a few.
While some of the Semitic languages have lost the emphasis distinction, Arabic still
retains the full set of emphatics reconstructed from Proto-Semitic at a number of places
of articulation (Versteegh, 2001; Watson, 2002). All dialects of Arabic have minimal
pairs of coronal obstruents where the difference is only in terms of emphatic vs. plain.
The present study uses data obtained from Urban Jordanian Arabic (UJA). It presents
an investigation of the acoustic and perceptual correlates of emphasis in UJA.
1.2. Emphasis
Different definitions of emphasis agree that it involves a posterior articulation. Studies
that used different dialects highlighted certain aspects of the definition of emphasis.
Among these, Walter Lehn provides a satisfactory definition of emphasis that
encompasses several aspects of the phenomenon. He says that:
Emphasis is the cooccurrence of the first and one or more others of the following
articulatory features: (1) slight retraction, lateral spreading, and concavity of the tongue
and raising of its back (more or less to what has been called velarization), (2) faucal
and pharyngeal constriction (pharyngealization), and (3) slight lip protrusion or
rounding (labialization), and (4) increased tension of the entire oral and pharyngeal
2
musculature resulting in the emphatics being noticeably more fortis than the plain
segments.
(Lehn, 1963:30-1)
Lehn’s definition accounts for more than one type of emphasis at the level of
articulation.
Lehn
justifiably
makes
a
distinction
between
emphasis
and
pharyngealization proper. Though most emphatics are pharyngeal, it has been found
that some emphatics are labialized in addition to their pharyngealization (Watson,
1999). Other definitions of emphasis are more restricted to pharyngealization. Some
Arab grammarians refer to emphasis as ½itbãq (literally, 'covering') and define it as
“spreading and raising of the tongue” (Lehn, 1963: 29). Others use the word ½isti¿laa½
(literally, 'elevation') and define it as “elevation of the dorsum,” ibid. Delattre (1971:
129) describes the production of pharyngeals as one in which “the root of the tongue
assumes the shape of a bulge and is drawn toward the vertical back wall of the pharynx
to form a stricture. This radical bulge generally divides the vocal tract into 2 cavities:
one below extending from the stricture to the glottis, the other above extending from
the stricture to the lip.” This provides a detailed description of the mechanism of the
production of pharyngeal sounds but it does not adequately account for the production
of emphatics as it involves another secondary yet crucial articulation. Kahn (1975: 39)
defines emphasis as a “secondary pharyngeal articulation of certain consonants, usually
stops and fricatives.” She adds that the articulation of emphasis involves the organs
engaged in the production of a given sound in addition to a secondary pharyngeal
articulation. This means that any consonant in Arabic, more precisely in Cairene
Arabic, the dialect she studies, can be emphatic. Though obstruents are the most
3
common emphatics in Arabic, some studies conducted on Cairene Arabic show that
emphasis can be a property of laterals and rhotics, too (Cf. Ferguson, 1956). This is also
supported by later reports from other studies conducted on other dialects of Arabic.
Zavadovsky (1981) cited in Zemánek (1996) reports that in the Mauritanian dialect
each consonant can have its emphatic and non-emphatic counterpart.
McCarthy (1994: 38) states that emphasis studied in different dialect regions in the
Arab world is always characterized by a "constriction in the upper pharynx". He
distinguishes between these emphatics and pharyngealized consonants claiming that
while the former ones are purely emphatic, the latter ones should be called uvularized –
affected by another set of back segments, i.e., uvulars: /q, ¼, Ë/.
Davis (1995: 465) defines emphasis, which corresponds, as he posits, to
pharyngealization, as the phenomenon of producing sounds “with a primary articulation
at the dental/alveolar region and with a secondary articulation that involves the
constriction of the upper pharynx.” He also provides an account for bilabial emphasis
in Arabic, which adds bilabials to the class of possible emphatics, not confining it to the
dental/ alveolar place of articulation. Emphasis at the labial region was later reported by
Watson (1999: 289) in a dialect of Yemeni Arabic. She says that “emphasis has two
articulatory correlates: pharyngealization and labialization,” and reports some examples
from Yemeni Arabic though there is a clear lack of minimal pairs at the labial place of
articulation. The bilabial consonant mentioned here is the bilabial nasal /m/. This
4
segment does not have an intrinsically emphatic counterpart. There is a set of minimal
pairs reported in the literature where /m/ is emphatic in one case and plain in the other –
/ma:lak/ vs. /m¿a:lak/, 'your (masc. 2nd person) money, and 'what's wrong with you
(masc. 2nd person)?' respectively. Another segment that shows similar behavior is /l/ in
sentences like /walla:hu/ and /wal¿l¿ahu/, 'he appointed him' and 'by God,' respectively.
These occurrences are not common in Arabic, and therefore I will not discuss them in
detail.
One of the most detailed studies of emphasis has been conducted by Card (1983). She
discusses emphasis from the phonological and phonetic perspectives in addition to
describing the articulation of emphasis. Rather than giving a brief definition of
emphasis, she provides a more detailed description of this phenomenon that I will leave
for the next section. More recent investigations on emphasis were provided by Laufer
and Baer (1988) who studied emphasis in Hebrew and Arabic, Zawaydeh (1999) who
studied gutturals in Arabic, Watson (2002) who found evidence for emphasis in uvulars
and bilabials, and Khattab, Al-Tamimi, and Heselwood (2006) who studied the gender
differences in the use of /t/ and /t¿/ in Jordanian Arabic. Their studies also fit better in
the next section. In summary, all previous definitions stress that the production of
emphatics involves a set of features that engage the pharynx in a mechanism much
similar to the production of pharyngeals.
5
1.3. Urban Jordanian Arabic (UJA)
One of the interesting debates in the literature is concerned with the linguistic situation
in Arabic. It is common to engage in discussions about the relation between Modern
Standard Arabic and the different dialects, and also the relation between these dialects
themselves and the different divisions of dialectal regions, hence the need to give a
clear idea about the dialect of Arabic from which the data for this study were collected.
The question of ‘what dialect of Arabic are you studying?’ has for a long time been of
interest to linguists. The linguistic situation within the Arab world is easy to describe:
each Arab community speaks its own dialect and understands a host of other Arabic
dialects including MSA, and to a lesser extent Classical Arabic (CA). Due to the wide
distribution of Egyptian movies and sitcoms in the Arab world, Egyptian Arabic seems
to be the most widespread dialect of the Arab world. Within each country, there are
distinct sub-dialects. Differences between these sub-dialects are more often than not
related to dissimilarity in the consonantal rather than the vocalic inventory. These
distinctions are usually hard to detect. For example, it is hard for a non-Moroccan
speaker of Arabic to distinguish different Moroccan dialects, to which he/she will
collectively refer as ‘Moroccan Arabic’. Similarly, Jordanian Arabic (JA) is commonly
spoken by Jordanians; however, there are different sub-dialects within what is
commonly referred to as JA that are hardly known to non-Jordanians. One of the
relatively vivid investigations of JA which addresses the linguistic situation in Jordan is
Suleiman (1985). He provides an investigation of the linguistic varieties, the
classical/colloquial dichotomy referred to as diglossia, that dominate the scene in
6
Jordanian Arabic, focusing on the linguistic practices of the educated groups in Jordan.
He collected data from 40 students at Yarmouk University in Jordan by means of
personal interviews and questionnaires. He found that the linguistic situation in Jordan
is best described as triglossic consisting of Classical Arabic, Colloquial Arabic and
MSA (p. 93). He divides JA into three dialectal regions: the ‘Madani,’ represented in
the urban centers, the ‘Fallahi,’ spoken in rural areas, and the ‘Bedouin,’ spoken by the
non-sedentary nomads, (p 13). These dialectal regions of JA differed in the consonantal
inventory. A more recent description divides JA into four sub-dialects: Urban, Rural,
Bedouin and Ghorani, (Zuraiq and Zhang, 2006). UJA is spoken by people living in the
major cities, including the capital city, the Rural dialect is spoken by villagers, mostly
from the northern villages of Jordan, the Bedouin dialect is spoken by the Bedouins in
the north east, middle and south of Jordan. And finally, the Ghorani dialect is spoken
by farmers and people living in the Jordan valley. As mentioned earlier, these dialects
differ in their consonantal inventories while the vowels are essentially the same. The
consonant inventory of UJA is shown in Table 1.
7
Table 1. Consonant inventory of Urban Jordanian Arabic (Zuraiq and Zhang, 2006)
ð s
ð¿ s¿
ß
k
g
x
¥
Glottal
}
Pharyngeal
Affricate
Approximant
Glide
f
Velar
Fricative
d
d¿
n
r
z
Palatal
m
t
t¿
Palatoalveolar
Nasal
Trill
Alveolar
b
Interdental
Labiodental
Labial
Plosive
½
š
¿ h
d¹
l
w
j
1.4. The articulatory and acoustic correlates of emphasis
As mentioned earlier, emphasis has been mostly studied from a phonological
perspective. However, many of the phonological observations might very well map on
to the acoustics of emphasis. Sibawayh, cited in Semaan (1968), described the
articulation of emphatic sounds as early as the eighth century AD. Sibawayh uses the
term ½itbãq, literally 'covering,' to refer to emphasis. He says that emphatic sounds are
produced by “the part of the tongue which is the place of their utterance being closely
covered in their utterance by what is opposite to it of the palate” Semaan (1968: 45). He
adds that in the production of emphatics “you raise it [the tongue] towards the palate.”
ibid. He also says that if it was not for emphasis, then /t¿/ would be /d/, /s¿/ would be /s/,
/z¿/ 1 would be /ð/ and /d¿/ would disappear from the language, (ibid p. 46). In other
words, Sibawayh suggests that the emphasis distinction creates minimal pairs differing
1
Semaan uses this phonetic symbol to refer to the emphatic voiced dental fricative.
8
in the feature [+emphatic] or [-emphatic]. This description pairs the /t¿/ with the /d/
rather than the /t/, where it should be, having in mind that the emphatic/ plain minimal
pairs agree in place and manner of articulation but differ only in emphasis. This account
also draws heavily on the physical observation of the speech organs that are brought
together in the production of emphatics. It should be noticed here that a clear account of
the status of these emphatics remains unavailable. Though Classical Arabic (CA), the
variety of Arabic spoken during the early Islamic era, retains the same orthography as
Modern Standard Arabic (MSA), it has undergone many changes at the phonetic level.
It is due to these and other changes that we now have MSA. A simple match of the
orthographic symbols to the same phonetic realizations of MSA is likely misleading.
Another account in the medieval literature on emphatics is sketched by Card (1983). In
addition to her account of Sibawayh’s study, much similar to what Semaan has
presented, she discussed Ibn Sina (980 – 1037) – known in the western literature as
Avicenna. Ibn Sina was a physician and his account of these sounds is much affected
by his scientific background. As an example, as discussed by Card, (1983: 9), he
describes the production of the sounds /t¿/, /t/, /d/, saying that they have the same place
of articulation. “For /t¿/ the air is restricted by the larger part of the tongue tip and the
two sides of the tongue. A depression is formed in the center of the tongue, which
resonates as the air is forcibly driven out. The /t/ has similar articulation but only the
tongue tip restricts the air.” As for the sound /d/, Ibn Sina states that, “there is no
covering of the palate,” and “the air is not strongly restricted,” hinting probably at the
9
energy of voicing in the vocal folds. This, I think, is an insightful description of these
three phonemes in Arabic much valued at a time when very few tools, if any, were
available to provide more sophisticated descriptions. For other descriptions of emphatic
as well as plain sounds of Arabic, Ibn Sina uses terminology such as ‘air bubbles’ and
‘yielding membranes,’ ibid, which are not as linguistic as those used by Sibawayh, yet
his observations are still linguistically valid.
These two accounts that have been presented give us an idea about emphasis but the
descriptions of the sounds might not be totally valid to account for the same sounds that
are now present in different dialects of Arabic. Sibawayh’s and Ibn Sina’s descriptions
cover the era from the eighth to the tenth century A.D. and possibly for a short time
after that. However, in light of the inevitable change that affects many sounds of the
world's languages over time, descriptions of certain sounds have to differ by extents
compatible with the degree of change.
Modern studies of emphasis have changed their focus from descriptions solely based on
observation to more detailed and accurate explanations made possible by available
techniques. Marcáis, as discussed by Card (1983: 13) investigated the articulation of
emphatics in the 1940s making use of palatograms and radioscopy of the vocal tract. He
found that the articulation of emphasis involves “muscular tension and retraction of the
prominent articulating organs.” “The tongue root approaches the back of the pharynx,
with the result that the back of the tongue forms a ‘collapsed plateau’, dipping away
10
from the palate.” The feature that holds for all descriptions of the production of
emphatics so far is that they all involve upper pharyngeal constriction. A distinction has
to be made here between pharyngeals and pharyngealized segments. The term
pharyngeals refers to the sounds whose primary articulator is the pharynx whereas the
term pharyngealized, used to describe emphatics, means that the pharynx is the
secondary articulator for sounds articulated primarily with other speech organs. Ghazeli
(1977) shows the distinction in the production of pharyngeals and pharyngealized
segments. Using cinefluographic films, he shows that for pharyngeals such as /¿/ and
/š/ the greatest constriction was located below the epiglottis, while the greatest
constriction for emphatics occurs in the upper pharynx. This suggests that the
production of these two types of sounds is unrelated – giving no legitimacy for using
the two terms interchangeably as some linguists do.
Lehn (1968: 31) has noticed other features of emphasis. In his phonological study, he
mentions that features of emphasis do not hold for speakers of different dialects to the
same degree. He observed that in Cairene Arabic, emphasis is more characteristic of
men than women and that effects of emphasis on adjacent segments, i.e. emphasis
spread, are also not similar. He does not, however, provide experimental support to
validate his statements that are built on articulatory impressions. If it is true that women
emphasize less than men do, we still need to know if this occurs in all environments
and in all dialects. These observations are mostly impressionistic.
Another
characteristic of emphasis that has been frequently observed and has experimental
11
support is that emphatics affect adjacent vowels by dragging them to the low back
position, hence lowering their F2 frequencies. With this in mind, Lehn’s observation
that “all plain consonants occur before plain vowels” ibid is straightforward. To that he
adds that emphasis never occurs as the only feature of any segment in any environment;
“its minimum domain is CV”. Lehn is the first one to establish that the minimum
emphasis domain is the syllable. He states that the minimum emphasis domain is the
syllable and that the maximum domain, later referred to as emphasis spread, is the
utterance. Further, he says that emphasis domain is calculated at the syllable level,
which means that a whole syllable is either emphatic or plain, (p. 37). For this last
argument, Lehn cites examples such as /t¿i:n/ vs. /ti:n/, 'mud' and 'figs', respectively;
and /z¿u:r/ vs. /zu:r/, 'perjury' and 'visit' (imperative), respectively. For these pairs,
subsequent studies (Card 1983, Davis 1995, and Watson 1999) report that the long
vowels block emphasis spread and are phonologically opaque to emphasis. However,
vowels blocking emphasis do not violate the minimum emphasis domain requirement.
In light of the fact that the present study is devoted to the acoustics of emphasis and
perception of emphasis, it remains unknown whether phonological opacity applies to
phonetic correlates.
Kahn (1975) compared the production of emphatics among males and females. She
found that emphatics in Cairene Arabic do not lower F2 frequency values to the same
degree for male and female speakers. Instead, the extent of F2 lowering for adjacent
segments, an indication of emphasis, is greater for male speakers than for female
12
speakers. This suggests that male speakers ‘emphasize’ more than female speakers,
seemingly an acoustic replication of Lehn’s (1968) phonological observation. Kahn
also tested a number of non-Arab speakers of American English who had been trained
by an Arab male speaker to produce emphatics and found that there was no significant
gender difference in their articulation of emphatic sounds. She concluded that “among
Arabs, the magnitude difference in the formant frequencies between men and women
was found to be much greater than anatomically predicted by Fant and the direction of
sex-related formant differences was opposite to Fant’s predictions.”
(p. 38). The
finding that differences in F2 frequencies between emphatic and plain segments were
smaller for native female than male speakers of Arabic (p.42) should be attributed to
sociolinguistic factors. If these patterns were due to anatomical properties, we would
expect both American and Cairene female speakers to emphasize less than males. The
fact that this was not the case suggests that both male and female speakers can
emphasize by similar degrees but don’t for sociolinguistic reasons: softening harsh
sounds, avoiding course sounds in favor of more ‘feminine’ sounds. That is, female
speakers intentionally emphasize less than males. Similar findings have been reported
in later research: Royal (1985: 155 - 157) on Gamaliya, a dialect of Egyptian Arabic,
and Haeri (1996: 70 - 74) also on Cairene Arabic. However, results contrary to the ones
mentioned above have also been reported. Al-Masri & Jongman (2004) found that
females had a greater degree of F2 lowering in emphatic segments, meaning that they
‘emphasized’ more than male speakers did.
13
Davis (1995) examined two dialects of Palestinian Arabic: a northern dialect and a
southern dialect. His analysis is phonologically motivated. He considers applications of
Grounded Phonology to account for emphasis spread in these two dialects. Davis
reports that emphasis spread behaves asymmetrically: leftward emphasis is not blocked
for the entire word but rightward emphasis is always blocked by [+high, –back] vowels
(p. 468). This goes against what Lehn suggested, namely that the vowels /i:/ and /u:/
always block emphasis spread, and it actually excludes the vowel /u:/ from being
opaque.
Davis' findings support the assumption that different dialects show asymmetries in the
characteristics of emphasis spread with respect to the so-called opaque segments that
block emphasis spread, and the direction of spreading. In more recent phonological
accounts put forward by Davis (1995) and Watson (1999), emphasis is realized by the
spread of the feature Retracted Tongue Root [RTR] to adjacent segments. This, I think,
maps on to the phonetic parameter of low F2. This mapping, however, might very well
be non-linear. In other words, the same distance of spread of the feature [RTR] in the
underlying representation might not correspond to a comparable lowering of F2 values
of the same segments in the same directions.
Card (1983) divides her investigation of emphasis into two parts: phonetic and
phonological. Having established a relatively clear account of the phonological
attributes of emphasis, I shall focus on the phonetic part of her study. Card studied
14
emphasis in Palestinian Arabic and found evidence for the lowering effect of emphasis
on F2 frequency for segments in emphatic environments compared to higher
frequencies for corresponding segments in plain environments. In addition to this bythen clear characteristic of emphasis, she found that emphasis spreads phonetically
rightward and leftward in the whole word, (p.49). Furthermore, she found that if the
emphatic segment is word-initial, then emphasis is likely to spread to an adjacent word.
Secondary emphatics, i.e. those that acquire emphasis effects from neighboring
segments through spread, do not have F2 frequencies as low as those that have
segments with primary emphasis or as high as F2 frequencies of plain segments. Card
noticed that emphasis spread is blocked in the presence of the vowels /i:/ and /u:/. The
presence of these vowels in a word eliminated the main effect of emphasis, F2 drop.
Interestingly, she reports that for one of the stimuli, /t¿i:r/, meaning ‘fly!’, she found a
significant F2 drop despite the occurrence of a ‘blocking’ segment (p. 79).
Royal (1985) explored the sociolinguistic factors crucial in linguistic choice. She
particularly investigated this choice in terms of the degree pharyngealization. Royal
observed a general tendency among Cairene women to have markedly less
pharyngealization than Cairene men. But with a closer look at her results, it is evident
that this difference is ascribed to conventional choice rather than anatomical constraints.
She found that older Gamalyian women, women of Gamalyia – a suburb of Cairo,
display stronger pharyngealization than older Gamalyian men (Royal, 1985:165). This
led Royal to question any anatomical constraints on the vocal tract configurations for
men and women as far as their capability of pharyngealization is regarded. Another
15
observation is that she noticed variability in the degree of pharyngealization among
Gamalyian subjects in response to the sex of the listener. In other words, she found that
the degree of emphasis varied depending on the gender of the addressee.
These
findings contradict Kahn’s (1975). Similar findings, that is, differences ascribed to
social norms rather than anatomical facts, were reported in the other studies mentioned
above.
Another study was conducted by Laufer and Baer (1988) on emphasis in Hebrew and
Arabic. They collected 300 minutes of video recordings and simultaneous audio
recordings for nine Hebrew and Arabic speakers using a fiberscope in the upper
pharynx. Their analysis shows that all the emphatic and pharyngeal sounds were made
with qualitatively the same pharyngeal constriction but that the pharyngeal
constrictions were more extreme for pharyngeal sounds than for oral emphatics. This
provides more evidence for the fact that emphatics and pharyngeals should be
distinguished, cf. McCarthy (1994).
Zawaydeh (1999) studied the phonology and phonetics of gutturals in a number of
Arabic dialects. She distinguishes two types of gutturals but reports that they behave
similarly. These two types of sounds are uvulars: /q, ¼, Ë/ and emphatics: /d¿, t¿, ð¿, s¿/.
Most of her work is devoted to studying the uvular sounds in terms of their
phonological and phonetic behavior. Zawaydeh supports McCarthy’s (1994) argument
that emphatics are uvularized sounds, neither velarized nor pharyngealized. Previous
studies (Ali and Daniloff 1972; Al-Ani 1978) showed that emphatics can be
16
pronounced with a secondary articulation either in the pharynx or at the velum.
Zemánek (1996) reports much similar findings.
Zawaydeh presents results on the acoustic characteristics of gutturals. She posits that
gutturals can be distinguished as one group consisting of emphatics, uvulars,
pharyngeals and laryngeals. Her findings indicate that gutturals show similar but not
identical acoustic behavior. For example, she reports that while pharyngeals are
characterized by a raising of F1 for adjacent vowels, emphatics show a strong effect of
F2 lowering for adjacent vowels. While effects of emphasis and uvularization can be
blocked, the segments that cause blocking are not similar for both classes. She also
distinguishes pharyngeals /š/ and /¿/ from emphatics in that pharyngeals do not cause
F2 drop for adjacent vowels but only a significant raising for F1. Further, her findings
show that F2 lowering spreads to adjacent syllables if it is triggered by emphatics not
by pharyngeals. This is evident in UJA in that low vowels after emphatics are
pronounced with a back allophone that does not appear after pharyngeals (p. 38 – 9).
Zawaydeh also questions whether the spread of emphasis and uvularization is
categorical or gradient. In other words, whether the F2 lowering or F2 raising of the
vowel in the target syllable and F2 of adjacent vowels drop gradually or categorically.
Her findings show that the drop is gradual but some discrepancies are also reported.
More recently, Watson (2002) provides an acoustic account of emphasis. She says that
“the oral emphatics are typically marked by a compact acoustic spectrum through
17
lowering of the upper frequency formant (principally F2) due to an enlarged mouth
cavity, and a raising of F1 due to a reduced pharyngeal cavity.” (p. 270). She does not
mention the effects of opaque vowels or the domain of emphasis.
Khattab, Al-Tamimi, and Heselwood (2006) investigated the gender differences in the
production of /t/ and /t¿/ in Jordanian Arabic. 10 speakers of JA (5 males and 3 females
were from the northern part of Jordan, 2 females were from the capital city of Amman)
produced 4 pairs of /t/ and /t¿/ in word-initial positions with the vowels /i/ and /a/. They
report lower F2 and higher F1 for vowels in emphatic environments compared to their
plain counterparts regardless of gender, (p. 151). They also report longer VOT delays in
/t¿/ than in /t/ only for females. They conclude that females generally have a markedly
less emphatic speech compared to males. In other words, VOT latencies were more
pronounced in male speech. However, this result, as they later explain (p. 155) is
compromised by the fact that the two female speakers that were from Amman are the
ones that show less emphasis. This is in line with the findings of Royal (1985) on
Gamalyian Arabic.
Another acoustic measure pertaining to emphasis is the locus equation. Locus equations
are ‘straight line regression fits to data points formed by plotting F2 transitions along
the y axis and their corresponding midvowel nuclei along the x axis,’ Sussman,
McCafrrey, and Matthews (1991). They provide information about the place of
articulation for the consonants, and also about the degree of coarticulation. Slopes
generated by plotting the F2 Onset by F2 Vowel indicate the degree of coarticulation.
18
Yeou (1997) investigated locus equations for a set of emphatic consonants: /d¿/, /t¿/,
/ð¿/, /s¿/ in Modern Standard Arabic spoken by native speakers of the Moroccan dialect.
He found that locus equations reflected some place distinctions. He also reports that the
pharyngealized set of consonants used in the study, which is the same set of primary
emphatics used in the present study, emerged as a distinct class, ‘having the flattest
locus equation slopes of all consonants.’ This points to minimal coarticulation between
the consonants and the vowels.
Another measurement is the spectral moments (center of gravity, standard deviation,
skewness and kurtosis). Norlin (1987) measured spectral moments for Egyptian
sibilants /s/ and /s¿/ in word-initial positions. He used FFT spectra created from a 26.5
ms sample taken after the first third of the sibilant, (p. 16). The target consonants he
used /s/ and /s¿/ were used in one vocalic environment /a:/.Norlin found that /s¿/ had a
significantly lower center of gravity (5974 Hz) compared to /s/ (6345 Hz), (p. 94).
Jongman, Wayland, and Wang, (2000) found that a greater positive skewness indicates
a concentration of energy in the lower frequencies.
Clearly, back articulation in several dialects of Arabic has triggered much research to
find out its phonological, articulatory, acoustic and, to a much smaller extent,
perceptual correlates. Many of these studies provide us with findings that help draw a
better picture of this back articulation in Arabic. Now we know more about the
mechanism of back articulation, spread of certain classes such as emphatics and
uvulars, blocking segments, directionality of spread and the importance of locus
19
equations. Despite these valuable findings, many of the studies in this respect suffer
from certain limitations. Some are too general to the point that they deal with a
relatively great number of back sounds despite the fact that the articulators of these
sounds are different. As suggested by Perturbation theory (Chiba and Kajiyama,1942),
not all back sounds have similar acoustic effects. Other limitations that are common
include using data from different dialect regions, combining a limited number of
subjects, and using data from the researcher himself/herself.
1.5. The perceptual correlates of emphasis
Perception of emphasis has received much less attention than the acoustics of emphasis.
Obrecht (1968) conducted a study on the perception of emphasis in Lebanese Arabic.
He synthesized a /t¿i:-ti:/ continuum in which he varied F2 in 120 Hz steps from 1080
Hz to 1800 Hz; F3 was fixed at 3000 Hz and F1 was flat for all the experiments.
Obrecht found that the perception of emphatics was categorical. The turning point
between emphatic and plain occurred at an F2 around 1560 Hz. Stimuli higher were
perceived as plain and stimuli lower were perceived as emphatic. He also found that the
perception of emphasis depended largely on vowel quality. In a replication of Obrecht’s
experiments for the /s¿i:-si:/ distinction in Moroccan Arabic, Yeou (1995) found similar
results. He also found that the F1 transition lacks perceptual cues for the plain/
emphatic distinction. Ali and Daniloff (1974) on Iraqi Arabic found that the vocalic
portion of the plain/ emphatic words included sufficient cues for listeners to detect the
distinction.
20
Other studies that report on the perception of emphasis viewed the results in
sociolinguistic terms related to gender, social selection, and existing linguistic norms
(Kahn 1975; Alwan 1983; Wahba 1993; and Al-Wer 2000b). Kahn (1975) states that
emphatics were equally perceived by native speakers of Arabic regardless of their
gender.
Due to the apparent scarcity of studies on the perception of emphasis, our
understanding of the characteristics of emphasis remains wanting. The previous studies
either focused on the emphatic/plain consonant, not on the vowel, or used the whole
word as the unit of perception. While focusing on the consonant gives us a good insight
into the nature of the target obstruents involved in emphasis, using the whole word falls
short of accounting for the source of the difference. Neither of them accurately
investigates the characteristics of the vowel adjacent to the target consonants.
1.6. Statement of the problem
The present study examines the acoustic and perceptual correlates of emphasis in UJA.
Many of the previous studies provided phonological accounts of emphasis including
emphasis domain, emphasis spread, directionality of emphasis spread, and opaque
vowels that delimit the domain of emphasis spread. Although several recent studies
explored the acoustic characteristics of emphasis, many aspects of the acoustic signal
21
for emphatic segments are still unclear. The relevance of the target consonants and
vowels and their effects on the acoustics and perception of emphasis is still debatable.
The present study focuses on the characteristics of the consonants by measuring their
duration, spectral moments and locus equations; and also on the vowels by measuring
their durations, F1, F2, and F3 at onset, temporal midpoint and offset of the vowel. The
present study also tests whether the findings on the acoustics of emphasis map on to
perception. That is, whether the perception of emphasis is realized most on the target
consonant or on the target vowel, with reference to some characteristics of the
consonants (manner) and the vowels (quality).
1.7. Rationale of the study
The current study examines the acoustics and perception of emphasis in UJA to provide
a better understanding of this phenomenon in Arabic linguistics. It aims to further
previous research on this matter and add deeper insights into the nature of emphasis.
Despite the increasing number of acoustic studies conducted on emphasis, several
factors contribute to the need for the present study. Much of the previous research had
limitations on the number of subjects and the inconsistency of their dialects, the nature
of the data used, the limited number of measurements, and in some cases the lack of
statistical analyses. With regard to the perception of emphasis, research is very scarce
to begin with. The limited number of studies on the perception of emphasis leaves this
aspect of emphasis virtually unexplored.
22
1.8. Objectives of the study
The present study aims at first better clarifying the nature of emphasis in UJA. While
many of the previous studies investigated emphasis in different dialects of Arabic, the
present one helps complete the picture by using data from a less commonly studied
dialect of Arabic. This should be helpful in enriching the different fields of Arabic
linguistics and possibly paving the way for much needed future research on Arabic.
One of the basic contributions of this study is to provide a better understanding of
emphasis through a systematic approach that enables the author to report on the
characteristics of the plain and emphatic segments by means of the measurements
conducted. The relatively fine control on the consonant types, using the different
vowels of UJA, using a relatively large and equal number of male and female subjects
and large number of observations should allow for reliable results. In addition to
contributing to the research on Arabic linguistics, a better understanding of the
characteristics of emphasis at the levels of acoustics and perception will definitely be
helpful for pedagogical purposes.
23
Chapter Two
Acoustic Study – the monosyllables
In this chapter, I present the first part of the acoustic study which deals with the
monosyllables. This part aims to introduce the basic findings on the acoustic correlates
of emphasis in UJA. Specifically, it aims to answer the following questions:
1.
What are the characteristics of the plain/emphatic consonants?
2.
Does manner of articulation of the consonants affect emphasis?
3.
Are the effects of emphasis equally attested for different vowels?
4.
How does emphasis spread within the syllable?
5.
Where is emphasis realized most?
6.
Does gender affect the degree of emphasis?
In accordance with perturbation theory (Chiba and Kajiyama, 1942), see also (Kent and
Read, 24:1992), it is expected that the presence of a constriction in the back part of the
vocal tract triggers F2 lowering and may raise F1 and F3. Many previous studies on
emphasis unanimously report F2 lowering of the vowel in the same syllable with the
target consonant (Al-Ani, 1970, 1978; Kahn, 1975; Ghazeli, 1977; Giannini & Pettorio,
1982; Card, 1983; Bukshaisha, 1985; Laufer & Baer, 1988; Haselwood, 1992;
Zawaydeh, 1999; Al-Masri & Jongman, 2004; Khattab, Al-Tamimi & Heselwood, 2006
among others). Fewer studies report F1 raising (Klatt & Stevens, 1969; Al-Ani, 1970,
1978; Ghazeli, 1977; Giannini & Pettorio, 1982; Alwan, 1986; Butcher & Ahmad, 1987;
24
Laufer & Baer, 1988). And finally, Card (1983) reports no differences between F3
values in plain and emphatic environments. In terms of durations, Obrecht (1968)
hinted that F2 consonant-vowel transitions were longer in emphatic positions compared
to their plain counterparts. In Egyptian Arabic, Norlin (1987:76) found that there was a
slight tendency for plain vowels to be longer than emphatic ones, especially for long
vowels. Alioua (1995), however, reported contrasting results, namely that vowels
following emphatic consonants had slightly longer durations than their plain
counterparts. And finally, Norlin (1987) found lower center of gravity for word-initial
emphatic fricatives compared to plain ones. Based on previous literature reviewed in
Chapter 1, I make the following hypotheses:
1.
Emphatic consonants will be slightly longer than their plain counterparts.
2.
Vowel duration is not affected by emphasis.
3.
Emphatic consonants will have a lower center of gravity than their plain
counterparts. They are also expected to have a greater positive skewness due to the
concentration of energy in the lower frequencies.
4.
Locus equations will show flatter slopes for emphatic consonants compared to
their plain counterparts.
5.
F2 will be significantly lower for the vowels adjacent to the emphatic consonant.
6.
F2 lowering will be strongest closest to the emphatic consonant.
7.
The presence of an emphatic consonant will raise F1 and F3 in adjacent positions.
8.
Gender does not affect emphasis; F2 lowering is expected to be similar for males
and females.
25
2.1. Method
2.1.1. Subjects
Eight adult speakers (four males and four females) of UJA aged 20 to 39 years old
(average age was 28.5 years) participated in the study. Three males, who were PhD
students at the University of Kansas, and their spouses were living on the KU campus;
the remaining male and female speakers lived in Fayetteville, AR. All of the subjects
spoke the urban dialect of JA. Duration of stay in the US at the time of recording
ranged from 3 months to 4 years (average duration of stay was 2.85 years). None of the
subjects reported any known speech or hearing impairments.
2.1.2. Materials and Procedure
A list of 96 target CVC stimuli was prepared. The wordlist consisted of 48 pairs of
words where the target consonant (the plain or emphatic consonant) was either wordinitial or word-final. Each of the UJA vowels was used with each of the target
consonants. Thus, four pairs of target consonants: /t/ - /t¿/, /d/ - /d¿/, /ð/ - /ð¿/, and /s/ /s¿/were used word-initially and word-finally along with the 6 vowels of UJA: /i:/, /i/,
/a:/, /a/, /u:/, and /u/ - yielding 48 words with plain consonants and 48 words with
emphatic consonants. The consonant /b/ was used to fill up the remaining C slot. Thus,
the study material consisted of CVb and bVC stimuli, see Appendix I. This yielded 48
nonwords (17 plain and 31 emphatic) that were all phonotactically acceptable in UJA.
Subjects read the target words in a carrier sentence:/'½iš.ki target word ka.'ma:n
'mar.rah/ [Say target word again]. Subjects read 3 times from a wordlist that included
26
12 pages, each having one column. Each column consisted of 12 sentences – the first
and last 2 sentences were fillers and did not include target words. Since Arabic short
vowels are not written in full letters, diacritics were added to illustrate these vowels
where they occurred. If subjects made a mistake while reading, they were instructed to
resume reading two lines before the line in which the mistake was made. Speakers read
the wordlist three times. Recordings were carried out in an anechoic chamber using a
Digital Master DAT recorder (Fostex D-5) and a unidirectional microphone
(ElectroVoice RE-20). Recordings were digitized at a 22050 Hz sampling rate using
Praat speech analysis software (Boersma and Weenink 2005). Each subject was
recorded individually and each recording session took approximately 12 minutes.
Measurements taken from the three repetitions were averaged for each subject and the
average was used for statistical analysis.
2.1.3. Measurements
For each consonant, duration was measured and spectral moments (center of gravity,
standard deviation, skewness and kurtosis) were calculated. Spectral moments were
calculated from FFT. All other measurements were taken from wide-band spectrograms.
For consonants in word-initial position, duration was measured from the offset of the
last segment of the carrier sentence, the vowel /i/, until the offset of the burst for stops
or the end of the frication duration for fricatives. For word-final stops, duration was
measured from the offset of the preceding vowel in the target word until the offset of
the stop burst. For word-final fricatives, the duration was that of the frication. For
fricatives, spectral moments were calculated with a 30-ms window centered around the
27
temporal midpoint of the fricative. For stop consonants, spectral moments were
calculated at the onset of the burst. As a result, window size for stop consonants varied
depending on the duration of the burst, which ranged from 10 to 46 ms. The burst
duration was longer, on average, for word-final stops compared to word-initial stops. In
both positions, the average burst duration was longer for voiceless stops compared to
voiced stops. In very few cases, the burst was very hard to locate. In these cases, the
burst was estimated using auditory and visual cues and a window of 10 to 15 ms,
depending on the duration of the burst, was placed over the estimated onset to calculate
the spectral moment. For the target vowels, duration was measured and F1, F2 and F3
were measured at onset, midpoint and offset of the vowel. The vowel onset is coded as
the point in the vowel closest to the onset of the target word and the vowel offset is the
point in the vowel closest to the offset of the target word regardless of the position of
the target consonant. In other words, in words with final target consonants, the
frequency measurements at the vowel offset resemble the point closest to the consonant
that triggers emphasis. Locus equations were also calculated for each consonant by
fitting a regression line to F2 values measured at the onset and the temporal midpoint of
each vowel where the target consonant was word-initial, or measured at the offset and
temporal midpoint of the vowel where the target consonant was word-final. Figure 1
shows an example of where the measurements were taken.
28
Spectral moments for fricatives
30ms
F3 onset
fricative duration F2 onset
F1 onset
F1, F2, F3 mid
F1, F2, F3 offset
burst: spectral
moments for
/b/ duration stops
Figure 1: Illustration of the nature and locations of acoustic measurements
2.2. Analysis
To analyze the data, different ANOVA tests were conducted on all measurements.
Bonferroni post hoc tests were also used. Twelve dependent variables and eight
independent variables were used, see Table 1 below.
29
Table 1: Dependent and independent variables used in the study
No.
1
Dependent variables
Target consonant duration
No.
1
2
Duration of /b/
2
3
4
Target consonant center of
gravity
Target consonant standard
deviation
Independent variables
Consonant-type (plain, emphatic)
Vowel quality (high front, low front,
high back)
3
Vowel length (long, short)
4
Target consonant position (initial, final)
5
Target consonant skewness
5
6
Target consonant kurtosis
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Vowel duration
Vowel F1 onset
Vowel F1 midpoint
Vowel F1 offset
Vowel F2 onset
Vowel F2 midpoint
Vowel F2 offset
Vowel F3 onset
Vowel F3 midpoint
Vowel F3 offset
Center of gravity for /b/
Standard deviation for /b/
Skewness for /b/
Kurtosis for /b/
7
8
Target consonant voicing (voiced,
voiceless)
Target consonant manner (stop,
fricative)
Word type (real word, nonword)
Gender (male, female)
2.3. Results
In this section, I present the results of the acoustic study for the monosyllables. I will
present the results on duration, formant frequencies, spectral moments and locus
equations. Since the target consonant appears either word-initially or word-finally, I
will divide the results accordingly. In all analyses, interactions between Consonant-type
30
and Gender never showed any significant differences. Except for one analysis, the same
applies for Word-type.
2.3.1. Duration
The results on target consonant durations in word-initial and word-final positions show
that emphatic consonants were consistently longer than their plain counterparts, [F (1,
766) = 4.747, p = 0.030]. This effect, however, seems to result from word-final target
consonants. Word-final emphatics are significantly longer than their plain counterparts,
[F (1, 382) = 5.681, p = 0.018], whereas word-initial plain and emphatic consonants are
not different in terms of duration, [F (1, 382) = 0.287, p = 0.592]. See Figure 2. The
presence of an asterisk (*) indicates significant differences.
140
Duration (ms)
120
*
100
* 102
96
110
Plain_all data
Emphatic_all data
101
92
94
Plain initial
Emphatic initial
Plain final
Emphatic final
80
60
Figure 2: Target consonant durations
31
When the target consonants are further divided by manner, results show that this
durational difference applies only to word-final stops. While word-final fricatives were
not significantly different, [F (1, 190) = 0.665, p = 0.416], word-final emphatic stops
were significantly longer than their plain counterparts, [F (1, 190) = 5.894, p = 0.016],
see Figure 3 below.
140
Duration (ms)
120
112
*
100
104
108
Plain stop
Emphatic stop
98
Plain fricative
Emphatic fricative
80
60
Figure 3: Target consonant duration in word-final positions by manner
No other significant main effects or interactions were found.
While the duration measurements show significant differences for the target consonants,
vowel durations were not affected by emphasis. Mean vowel duration was 132 ms for
vowels in plain environments and 134 ms for vowels in emphatic environments. No
significant main effects or interactions were found.
32
All target words either started or ended with /b/. An investigation of durational
differences of this consonant in the presence or absence of an emphatic segment yielded
no significant differences. For all tokens, /b/ duration was 84 ms and 85 ms in plain and
emphatic environments, respectively. Similar results for duration are maintained in
words with initial emphatics: 88 ms in plain environments and 89 in emphatic
environments. The same results are also obtained in words with final emphatics: /b/
duration is 80 ms in words with plain consonants and 81 ms in words with emphatic
consonants. No significant main effects or interactions were found.
2.3.2. Formant frequency
In this section, I report on the three vowel formant frequencies, F1, F2 and F3 in three
positions: vowel onset, midpoint and offset.
2.3.2.1. F1
Overall, vowels in emphatic environments have a higher F1 onset compared to vowels
in plain environments. Including measurements from all data, F1 onset is 397 Hz in
emphatic positions and 369 Hz in plain positions. This difference is significant, [F (1,
766) = 26.402, p < 0.001]. This difference holds for words where the emphatic is wordinitial, F1 onset is 403 Hz in emphatic environments and 364 Hz in plain environments,
[F (1, 382) = 25.196, p < 0.001]. The same comparison also yields significance in
33
word-final emphatics, F1 onset is 390 Hz for vowels in emphatic positions, and 372 Hz
for vowels in plain positions: [F (1, 382) = 5.194, p = 0.023], see Figure 4 below.
Frequency (Hz)
500
400
*
397
369
403
*
364
*
390
372
Plain_all data
Emphatic_all data
Plain initial
Emphatic initial
Plain final
300
Emphatic final
200
Figure 4: F1 at vowel onset
A Consonant-type by Position interaction reveals a trend. Though it does not show
significance, [F (1, 767) = 3.599, p = 0.058], there is a tendency for F1 at vowel onset
to rise by a larger extent in words with initial target consonants compared to words with
final target consonants: 364 Hz to 403 Hz for words with initial target consonants
compared to 372 Hz to 390 Hz for words with final target consonants; magnitudes of
rise are 39 Hz and 18 Hz, respectively. No other significant main effects or interactions
were found.
F1 at midpoint also shows significant differences. Across all data, vowels in the same
syllable with the emphatic target consonants have a significantly higher F1 at midpoint
34
than vowels in the same syllable with plain target consonants: 508 Hz and 480 Hz,
respectively, [F (1, 766) = 5.448, p = 0.020]. However, when the data are divided by
position, this difference is no longer significant. F1 midpoint for vowels in words with
initial emphatics is 509 Hz, and 483 Hz for their plain counterparts, [F (1, 382) = 2.315,
p = 0.129]. In words with emphatics in final position, F1 midpoint is 507 Hz in
emphatic environments, and 477 Hz in plain environments, [F (1, 382) = 3.160, p =
0.076]. See Figure 5 below.
600
Frequency (Hz)
500
508
*
480
509
483
507
477
Plain_all data
Emphatic_all data
Plain initial
400
Emphatic initial
Plain final
Emphatic final
300
200
Figure 5: F1 at vowel midpoint
No other significant main effects or interactions were found.
At offset positions, F1 also shows significant differences. F1 offset is higher in
emphatic positions than in their plain counterparts, 409 Hz and 381 Hz, respectively: [F
35
(1, 766) = 19.768, p < 0.001]. This difference is lost in words with initial target
consonants but is maintained in word-final target consonants. F1 offset is 409 Hz in
emphatic environments and 395 Hz in plain environments in words with initial
emphatics. This difference is not significant, [F (1, 382) = 2.071, p = 0.151]. In words
where the target consonant is word-final, F1 offset is 409 Hz in emphatic environments
and 366 Hz in their plain counterparts, [F (1, 382) = 26.181, p < 0.001]. See Figure 6.
Frequency (Hz)
500
400
409
*
381
*
395
409
409
*
366
Plain_all data
Emphatic_all data
Plain initial
Emphatic initial
Plain final
300
Emphatic final
200
Figure 6: F1 at vowel offset
A Consonant-type by Position interaction shows that while words with final target
consonants have a significantly higher F1 offset in emphatic environments (409 Hz)
compared to plain environments (366 Hz), words with initial target consonants lack this
distinction for F1 offset in emphatic positions (409 Hz) and in their plain counterparts
36
(395 Hz) [F (1, 767) = 5.365, p < 0.021]. See Figure 7 below. No other significant main
effects or interactions were found.
Frequency (Hz)
500
400
Word-initial
Word-final
300
200
Plain
Emphatic
Figure 7: Consonant-type by Position interaction for F1 Offset
Since the F1 measurements were taken at three positions, it is worthwhile to look at the
effects of emphasis across the three points of the vowel: onset, midpoint and offset. In
words with initial target consonants, F1 is significantly higher in emphatic
environments. This difference is significant only at the onset of the vowel – the point
closest to the consonant that triggers emphasis. The effect diminishes as the
measurement is taken further away from the target consonant, see Figure 8. Arrows
stand for the direction of significant differences; absence of the arrows indicates
absence of significant differences.
37
Similarly, in words with final target consonants, vowel F1 at the point closest to the
target consonant (F1 offset in this case) is significantly higher in the emphatic
environment. The effect diminishes at the vowel midpoint, though it shows a trend, and
is significantly higher at the point further away from the target consonant, (F1 onset in
this case). See Figure 9.
Frequency (Hz)
600
509
483
500
Plain
Emphatic
400
409
395
403
364
300
Onset
Mid
Offset
Figure 8: Vowel F1 in words with initial target consonants
38
Frequency (Hz)
600
507
500
477
Plain
Emphatic
400
409
390
372
366
300
Onset
Mid
Offset
Figure 9: Vowel F1 in words with final target consonants
2.3.2.2. F2
Measurements for F2 across all data show that F2 is consistently lower in emphatic
than in plain environments. F2 measured at the onset of the vowel in the same syllable
with the target consonant shows significant lowering in the emphatic environment.
Average F2 onset is 1332 Hz in emphatic environments and 1704 Hz in plain
environments. This difference is significant, [F (1, 766) = 212.738, p < 0.001]. This
difference is significant in both word-initial and word-final emphatics. In words with
initial target consonants, F2 onset for vowels is 1332 Hz in emphatic environments and
1838 Hz in plain environments. This effect is significant, [F (1, 382) = 342.174, p <
0.001]. In words with final target consonants, F2 onset is 1341 Hz and 1569 Hz in
emphatic and plain environments, respectively. This is also significant, [F (1, 382) =
31.746, p < 0.001]. See Figure 10.
39
2000
1838
Frequency (Hz)
1800
1704
Plain_all data
Emphatic_all data
* 1569
1600
*
*
1400
1332
1322
1341
Plain initial
Emphatic initial
Plain final
Emphatic final
1200
1000
Figure 10: F2 at vowel onset
An interaction of Consonant-type by Position reveals significant differences [F (1, 767)
= 34.161, p < 0.001]. The drop in F2 onset is significantly greater in words with initial
emphatics (516 Hz) compared to words with final emphatics (228 Hz). Figure 11 shows
this result.
40
2000
1800
1600
Word initial
1400
Word final
1200
1000
Plain
Emphatic
Figure 11: Consonant-type by Position for vowel F2 onset
In addition, a Consonant-type by Vowel Quality interaction reveals significant
differences, [F (1, 767) = 15.544, p < 0.001]. While the effect of emphasis is similar for
the front vowels /i:/ and /i/ and /a:/ and /a/, the F2 decrease for the back vowels /u:/ and
/u/ is much smaller. Table 2 below shows these results. The same results are obtained
for words with initial target consonants, [F (1, 383) = 13.043, p < 0.001], and for words
with final target consonants, [F (1, 383) = 13.097, p < 0.001]. See Table 3 below. No
other significant main effects or interactions were found.
Table 2: F2 onset by vowel quality
High front
Low front
High back
Plain (Hz)
2069
1666
1376
Emphatic(Hz)
1627
1221
1146
41
Plain-emphatic
442
445
230
% difference
21
27
17
Table 3: F2 onset by vowel quality in word-initial and word-final target consonants
High front
Low front
High back
Plain (Hz)
Initial
Final
2131
2007
1769
1564
1616
1137
Emphatic (Hz)
Initial
Final
1486
1769
1240
1202
1241
1052
% difference
Initial
Final
30
12
30
23
23
7
Vowel F2 at midpoint shows a significant drop in emphatic environments. Vowels in
emphatic environments have an F2 midpoint value of 1428 Hz compared to 1678 Hz
for their plain counterparts, [F (1, 766) = 19.768, p < 0.001]. In words with initial target
consonants, F2 midpoint is still lower in emphatic environments, 1427 Hz compared to
their plain counterparts, 1680 Hz, [F (1, 382) = 26.784, p < 0.001]. The same effect is
also detected for words with final target consonants. F2 midpoint in words with final
target consonants is 1428 Hz in emphatic positions and 1676 Hz in plain positions, [F
(1, 382) = 21.837, p < 0.001]. See Figure 12 below.
42
2000
1800
Frequency (Hz)
1678
1600
1680
*
1676
Plain_all data
*
*
Emphatic_all data
Plain initial
1428
1427
1428
1400
Emphatic initial
Plain final
Emphatic final
1200
1000
Figure 12: F2 at vowel midpoint
No other significant main effects or interactions were found.
A significant Consonant-type by Vowel Length interaction reveals differences between
vowels in the two different conditions. Independent of emphasis, long vowels (164 ms)
are longer than short vowels (102 ms). This difference is significant: [F (1, 766) =
784.735, p < 0.001]. In short vowels, F2 midpoint is 1648 Hz in plain environments and
1321 Hz in emphatic environments, a drop of 327 Hz. In long vowels, F2 midpoint is
1708 for vowels in plain environment and 1534 Hz in emphatic environments, a drop of
174 Hz. This difference in lowering is significant: [F (1, 767) = 4.621, p = 0.032]. See
Figure 13 below.
43
2000
Short vowels
Long vowels
Frequency (Hz)
1800
1708
1648
1600
1534
1400
1321
1200
1000
Plain
Emphatic
Figure 13: Consonant-type by Vowel length for vowel F2 midpoint
However, this difference in F2 midpoint is not attested for words with initial target
consonants, [F (1, 383) = 2.468, p = 0.117], nor for words with final target consonants,
[F (1, 383) = 2.156, p = 0.143].
F2 midpoint is affected by vowel quality. Overall, low front vowels show the highest
degree of lowering for F2 midpoint between plain and emphatic environments,
followed by high front vowels. The least degree of lowering is detected for high back
vowels. These differences are significant, [F (1, 767) = 14.770, p < 0.001].Table 4
below shows these values.
44
Table 4: F2 midpoint by vowel quality
High front
Low front
High back
Plain (Hz)
2168
1669
1196
Emphatic (Hz)
1959
1254
1070
Plain - emphatic
209
415
126
% difference
10
25
11
The same effect is attested for words with initial target consonants, [F (1, 383) = 7.220,
p < 0.001], and also for words with final target consonants, [F (1, 383) = 9.159, p <
0.001]. Table 5 below shows these results.
Table 5: F2 midpoint by vowel quality in word-initial and word-final target consonants
High front
Low front
High back
Plain (Hz)
Initial
Final
2131
2206
1668
1670
1240
1151
Emphatic (Hz)
Initial
Final
1969
1949
1256
1253
1056
1083
% difference
Initial
Final
8
12
25
25
15
6
F2 offset is also significantly lower in emphatic environments. Across all data, F2
offset drops from 1615 Hz in plain environments to 1302 in emphatic ones, [F (1, 766)
= 164.548, p < 0.001]. F2 offset values demonstrate the same effects in words with
initial target consonants and words with final target consonants. In word-initial
positions, F2 offset is 1486 Hz in plain environments and 1321 Hz in their emphatic
counterparts, F (1, 382) = 19.613, p < 0.001]. In words with final target consonants, F2
offset is 1744 Hz in plain environments and 1282 Hz in their emphatic counterparts, F
(1, 382) = 258.281, p < 0.001]. Figure 14 shows these results.
45
2000
1744
Frequency (Hz)
1800
Plain_all data
1615
Emphatic_all data
1600
*
1400
1302
1486 *
1321
Plain initial
*
1282
Emphatic initial
Plain final
Emphatic final
1200
1000
Figure 14: F2 at vowel offset
An interaction between Consonant-type and Position of the target consonant reveals
significant results. F2 offset in words with final target consonants drops by 462 Hz,
(1744 in plain environments and 1282 in emphatic ones). In words with initial target
consonants, F2 offset drops only by 165 Hz, (1486 Hz in plain environments compared
to 1321 Hz in their emphatic counterparts). This difference proves significant: [F (1,
767) = 39.683, p < 0.001]. Another interaction between Consonant-type and Vowel
Quality reveals significant differences for F2 offset. High front vowels have the largest
F2 offset drop, followed by low front vowels and finally by high back vowels. These
differences are significant, [F (2, 767) = 9.223, p < 0.001]. Table 6 shows these results.
46
Table 6: F2 offset by vowel quality
High front
Low front
High back
Plain (Hz)
1948
1551
1346
Emphatic (Hz)
1569
1191
1145
Plain - emphatic
379
360
201
% difference
19
23
15
The same effects hold for words with initial emphatics where low front vowels show
the largest drop, followed by high front vowels and then high back vowels. These
differences are significant, [F (2, 383) = 8.981, p < 0.001]. In words with final
emphatics, high front vowels drop by the largest amount, followed by low front vowels
and finally by high back vowels, [F (2, 383) = 7.587, p < 0.001]. Table 7 below shows
these findings.
Table 7: F2 offset by vowel quality in word-initial and word-final target consonants
High front
Low front
High back
Plain (Hz)
Initial
Final
1875
2021
1448
1654
1135
1558
Emphatic (Hz)
Initial
Final
1702
1437
1168
1214
1092
1196
% difference
Initial
Final
9
39
19
27
4
23
No other significant main effects or interactions were found.
Plotting the results for F2 with regard to the point at which the measurements were
taken reveals interesting observations. The closer the measurement is taken to the
emphatic consonant, the larger the F2 drop in emphatic environments. In Figure 15, this
F2 drop is greatest at vowel onset, the point closest to the target consonant, and
47
decreases as the measurement point moves away from the target consonant. The mirror
image is observed in Figure 16 since the target consonant is word-final.
2000
Frequency (Hz)
1800
1838
1680
Plain
1600
Emphatic
1486
1427
1400
1332
1321
1200
Onset
Mid
Offset
Figure 15: Vowel F2 in words with initial target consonants
2000
Frequency (Hz)
1800
1744
1676
1600
Plain
1569
Emphatic
1428
1400
1341
1282
1200
Onset
Mid
Offset
Figure 16: Vowel F2 in words with final target consonants
48
2.3.2.3. F3
Overall, F3 in emphatic environments rises in all three positions: onset, midpoint and
offset, compared to plain ones. Across all data, F3 onset is 2811 Hz in plain
environments compared to 2880 Hz in emphatic environments. This difference is
significant, [F (1, 766) = 11.469, p = 0.001]. The same difference is attested for words
with initial target consonants: F3 onset is 2895 Hz in plain environments and 3005 Hz
in emphatic environments, [F (1, 382) = 15.892 p < 0.001]. However, F3 onset does not
show significant differences in words with final target positions. It is 2727 Hz in plain
environments compared to 2756 Hz in emphatic environments, [F (1, 382) = 1.233, p =
0.268]. Figure 17 shows these values.
3200
3000
* 3005
*
2880
Frequency (Hz)
2811
2800
2895
2727 2756
Plain_all data
Emphatic_all data
Plain initial
2600
Emphatic initial
Plain final
2400
Emphatic final
2200
2000
Figure 17: F3 at vowel onset
49
An interaction between Consonant-type and Word-type reveals significant differences
for F3 onset. In real words, F3 onset is higher in emphatic positions (2928 Hz) than in
plain positions (2810 Hz) by 118 Hz. In nonwords, this difference in F3 onset is only
28 Hz: 2840 Hz for F3 onset in emphatic nonwords compared to 2812 Hz in their plain
counterparts. This difference is significant. See Figure 18.
3200
Frequency (Hz)
3000
Word
2800
Nonword
2600
2400
Plain
Emphatic
Figure 18: Consonant-type by Word-type for vowel F3 onset
No other significant main effects or interactions were found.
F3 at vowel midpoint is higher in emphatic environments. When all data are included,
vowels in emphatic positions have an F3 midpoint of 2937 Hz compared to 2870 Hz in
plain positions; [F (1, 766) = 8.610, p = 0.003]. In words with initial target consonants,
F3 midpoint is still higher in emphatic environments, 2945 Hz, compared to 2869 Hz in
their plain counterparts, [F (1, 382) = 5.373, p = 0.021]. In words with final target
50
consonants, F3 midpoint is 2929 Hz in emphatic environments and 2870 Hz in their
plain counterparts. Though F3 midpoint is still higher in emphatic positions, this effect
is not significant, [F (1, 382) = 3.331, p = 0.069]. Figure 19 shows these results.
3200
Frequency (Hz)
3000
* 2937
2870
* 2945
2929
2869
2870
Plain_all data
2800
Emphatic_all data
Plain initial
2600
Emphatic initial
Plain final
2400
Emphatic final
2200
2000
Figure 19: F3 at vowel midpoint
A Consonant-type by Vowel Quality interaction for F3 midpoint proves significant: low
front vowels have a higher F3 midpoint in emphatic positions than in plain ones while
high front vowels have a lower F3 midpoint in emphatic positions compared to plain
ones, [F (2, 767) = 5.367, p = 0.005]. However, F3 midpoint in high back vowels does
not show significant differences. Table 8 below shows F3 midpoint values. The minus
symbol represents the direction of difference – namely that the value is larger in
emphatic positions.
51
Table 8: F3 midpoint by vowel quality
High front
Low front
High back
Plain (Hz)
2971
2821
2816
Emphatic (Hz)
2958
2871
2981
Plain - emphatic
13
-50
-165
% difference
0.5
-1
-1
When data are split by position of the target consonant, this interaction is not observed
for words with initial target consonants. There is no significant interaction between
consonant-type and vowel quality for F3 midpoint in words with initial target
consonants, [F (1, 383) = 1.249, p = 0.288]. The magnitude of drop for F3 midpoint is
significantly different for high front vowels compared to low front vowels, where F3
rises in emphatic positions, whereas high back vowels do not show significant
differences. Table 9 shows these values.
Table 9: F3 midpoint by vowel quality in word-initial and word-final target consonants
High front
Low front
High back
Plain (Hz)
Initial
Final
2953
2990
2825
2818
2830
2801
Emphatic (Hz)
Initial
Final
2968
2948
2897
2845
2970
2992
% difference
Initial
Final
-1
0.2
-1
-1
-1
-1
No other significant main effects or interactions were found.
Across all data, F3 offset is higher in emphatic positions. F3 offset is 2917 Hz for
vowels in emphatic positions, and 2838 Hz for vowels in plain positions, [F (1, 766) =
12.467, p < 0.001]. For words with initial target consonants, F3 offset is 2812 Hz in
emphatic positions, and 2752 Hz in plain positions. This effect shows a trend but is not
52
significant, [F (1, 3.652) = 8.610, p = 0.057]. For words with word-final target
consonants, F3 offset is 3021 Hz in emphatic positions compared to 2923 Hz in their
plain counterparts. This difference is significant, [F (1, 382) = 11.535, p = 0.001].
Figure 20 plots these results.
3200
3000
*
* 3021
2923
2917
Frequency (Hz)
2838
2800
2752
2812
Plain_all data
Emphatic_all data
Plain initial
2600
Emphatic initial
Plain final
2400
Emphatic final
2200
2000
Figure 20: F3 at vowel offset
No other significant main effects or interactions were found.
When looking at the change in F3 measured at different positions of the vowel, results
are compatible with what is reported for F1 and F2 above. The closer the measurement
is taken to the consonant that triggers emphasis, the greater the rise. As the
53
measurement is taken further away, this rise either declines on disappears. See Figure
21 and Figure 22.
Frequency (Hz)
3200
3000
3005
2895
2945
Plain
2869
Emphatic
2812
2800
2752
2600
Onset
Mid
Offset
Figure 21: Vowel F3 in words with initial target consonants
Frequency (Hz)
3200
3021
3000
2929
2923
2870
Emphatic
2800
2756
2727
2600
Onset
Plain
Mid
Offset
Figure 22: Vowel F3 in words with final target consonants
54
2.3.3. Spectral moments
2.3.3.1. Target consonants
Overall, none of the spectral moments for plain and emphatic consonants show
significant differences. Center of gravity was 6840 for plain consonants and 6740 for
emphatic consonants, [F (1, 766) = 0.999, p = 0.318]. The same results are obtained for
word-initial emphatics – center of gravity is not different between plain and emphatic
consonants. A Consonant by Manner interaction reveals significant differences. In
word-final target consonants, plain stops have a higher center of gravity (6793 Hz)
compared to their emphatic counterparts (6485 Hz), whereas plain fricatives have a
lower center of gravity (7906 Hz) compared to their emphatic counterparts (7945 Hz).
This difference is significant, [F (1, 383) = 4.082, p = 0.044]. See Figure 23 below. No
other significant main effects or interactions were found.
8000
7945
7906
Frequency (Hz)
7600
7200
6800
Stop
Fricative
6793
6485
6400
6000
Plain
Emphatic
Figure 23: Consonant-type by Manner of articulation for Center of gravity in words with final
target consonants
55
For standard deviation, no significant main effects or interactions were found. Mean
variance was 2330 Hz for plain consonants and 2359 Hz for emphatic consonants, [F (1,
766) = 0.255, p = 0.613].
No significant main effects or interactions were found for skewness. Plain consonants
had a skewness value of 3.464 and -0.4711 for their emphatic counterparts, [F (1, 766)
= 1.000, p = 0.318].
Finally, for kurtosis, no significant main effects or interactions were found. Plain target
consonants have a kurtosis value of 0.3696 and 0.5755 for their emphatic counterparts,
[F (1, 766) = 2.214, p = 0.137].
2.3.3.2. /b/ consonant
Overall, spectral moments for /b/ yielded no significant main effects or interactions.
Table 10 shows these results.
Table 10: Spectral moments for /b/
Measurement
Center of gravity (Hz)
Standard deviation (Hz)
Skewness
Kurtosis
Plain
5418
3258
0.0067
-0.9954
Emphatic
5405
3277
0.0035
-0.9946
Similar results are obtained for words with initial target consonants and words with
final target consonants as well, see Table 11.
56
Table 11: Spectral moments for /b/ in word-initial and word-final target consonants
Measurement
Center of gravity (Hz)
Standard deviation (Hz)
Skewness
Kurtosis
Plain
Initial
5922
3029
-0.1431
-0.9582
Final
4941
3488
0.1566
-1.0326
Emphatic
Initial
Final
5952
4859
3035
3520
-0.1764
0.1834
-0.9580
-1.0289
2.3.4. Locus equations
Locus equations were calculated for each target consonant. A linear regression was
conducted with F2 onset as the dependent variable and F2 vowel as the independent
variable to obtain the slope and the y-intercept values. A one-way ANOVA shows that
emphatic target consonants have a lower mean slope (0.464) than their plain
counterparts (0.640). This difference is significant, [F (1, 14) = 5.586, p = 0.033]. This
effect is attested for words with initial target consonants: slope value is 0.310 for
emphatic consonants, and 0.576 for their plain counterparts, [F (1, 6) = 11.223, p =
0.015]. The same effect is also evident for words with final target consonant: slope
value is 0.618 for emphatic consonants and 0.708 for their plain counterparts, [F (1, 6)
= 7.153, p = 0.037]. See Figure 24. Figure 25 shows the scatter plots for all the data.
57
2
Plain_all data
Slope
Emphatic_all data
Plain initial
1
*
0.64
*
0.576
*
0.708
0.618
0.464
Emphatic initial
Plain final
Emphatic final
0.31
0
Figure 24: Slope values
58
1:plain; 2:emphatic
plain
emphatic
3000.00
F2 Onset (Hz)
2500.00
2000.00
1500.00
1000.00
R Sq Linear = 0.652
500.00
R Sq Linear = 0.585
0.00
0.00
1000.00
2000.00
3000.00
F2 Midpoint (Hz)
Figure 25: scatter plot, all data
Analysis of y-intercept values did not reveal significant differences. For all data, mean
y-intercept was 630 Hz for plain and 668 Hz for emphatic consonants, [F (1, 14) =
0.075,
p = 0.788]. The same results were obtained for words with initial target
consonants. y-intercept value was 872 Hz for plain consonants, and 878 Hz for their
emphatic counterparts, [F (1, 6) = 0.002 , p = 0.962]. Similarly, words with final target
consonants show no difference in y-intercept values. Plain consonants have a y-
59
intercept of 388 Hz and their emphatic counterparts have a y-intercept of 457 Hz, [F (1,
6) = 3.252, p = 0.121]. Figure 26 below shows these results.
1000
872 878
Frequency (Hz)
800
Plain_all data
630
668
Emphatic_all data
Plain initial
600
Emphatic initial
457
400
388
200
Figure 26: y-intercept value
60
Plain final
Emphatic final
2.4. Results Summary
Tables 12 and Table 13 summarize the findings of the acoustic study. This arrow (y)
represents an increase in emphatic positions, the other one (z) represents a decrease in
the emphatic position, and an arrow with a (t) represents a trend in the direction of the
arrow. The number of arrows in one box indicates the magnitude of the raise or drop.
Variables that did not show significant differences are excluded from these two tables.
Table 12: Main findings
C duration
F1 onset
F1 midpoint
F1 offset
F2 onset
F2 midpoint
F2 offset
F3 onset
F3 midpoint
F3 offset
All
y
y
y
y
z
z
z
y
y
y
Initial
y
z
z
z
y
y
ty
61
Final
y
y
ty
y
z
z
z
y
Table 13: Summary of interactions: from left to right: I: Initial; F: Final; S: Stop; F: Fricative;
VD: Voiced; VS: Voiceless; S: Short; L: Long; W: Word; N: Nonword; M: Male; F: Female.
Position
I
C duration
Manner
F
S
y
y
F
Voicing
VD VS
V length
S
L
V quality
/i/
/i:/
/a/
/a:/
/u/
/u:/
W N
M
F
ty
F1 midpoint
F2 onset
Gender
y
/b/ duration
F1 onset
Wordtype
ty
z
z
F2 midpoint
F2 offset
z
zz
zz
z
zz
zz
z
z
zz
z
zz
z
y
F3 onset
z
F3 midpoint
C gravity
y2
y
z3
2.5. Discussion
The present results show that emphatic stops in word-final position are consistently
longer than their plain counterparts. Preliminary results (Al-Masri and Jongman, 2004)
indicated an effect of position of the target consonant – namely, that word-initial target
emphatics tend to have longer durations than their plain counterparts. However, this
tendency might have been influenced by the fact that it was obtained from data with
2
3
Only in word-final position
Only in word-final position
62
different word lengths. The results obtained here suggest that stops are more sensitive
to emphasis than fricatives in terms of duration. This might be due to the fact that, in
general, fricative durations are longer than stop durations, which means that stops are
more likely capable of showing slight durational differences compared to fricatives.
Besides, this supports Zawaydeh’s (1999) finding about the directionality of emphasis
in UJA, namely that it is more prominent leftward than rightward. The lack of such
findings for the other segments, the vowel and the /b/ consonant, might be due to the
fact that durational differences in emphasis are not likely to spread in either direction.
While durational differences are observed on the target consonants under certain
conditions, they are not expected to spread to adjacent segments. Ultimately, back
articulation, which is characteristic of emphasis, is not closely related to duration.
As for formant frequencies, different results are reported. The overall means do not
seem to be incongruent with previous findings, (Obrecht, 1968; Alosh 1987; Zawaydeh
1999). However, the present results are more detailed since all formant frequencies
were measured at three positions of the vowel: onset, midpoint and offset. F1 onset is
consistently higher in emphatic than in plain positions. There is a trend for F1 to rise by
a larger amount in words with initial target consonants compared to words with final
consonants. F1 at the steady state of the vowel is still higher in emphatic positions
compared to plain ones for all data. This difference is lost for word-initial emphatics
and is only obtained as a trend for word-final emphatics. The lack of significance for F1
rise in the steady state of the vowel for word-initial target consonants, and the presence
63
of a slight trend for words with final target consonants support the previous findings in
this study for the prominence of leftward spread of emphasis within monosyllables. For
F1 at offset positions, a mirror image of F1 onset is presented. Overall, F1 offset is
higher in emphatic positions. Also, the magnitude of rise for F1 offset is clearer for
word-final target consonants. From the findings for F1 onset and F1 offset, we can
conclude that emphasis is gradient within the monosyllable: the closer the vowel and
the segment that triggers emphasis, the more pronounced the rise in F1.
Most of the previous studies report on F2. F2 is lower in emphatic positions than in
plain ones at the three measurements points. F2 onset is significantly lower for all data
and when the target consonant is word-initial and word-final as well. The difference
between F2 onset values in emphatic positions is significantly more pronounced in
word-initial than in word-final positions. Another detailed finding is that the drop is not
similar for different vowels. It is evident that the drop is larger for high front and low
front vowels compared to high back vowels. Seemingly, there is more room for this
drop in front vowels than in back vowels. Additional interesting results are observed for
F2 midpoint. Consistently, F2 midpoint is lower in emphatic environments.
Interestingly, no interaction with position of the target consonant is observed here now
that the measurements are taken equidistant from the consonant that triggers emphasis.
Interactions with vowel quality are also more salient here. Low front vowels show the
largest F2 midpoint drop, followed by high front vowels and finally high back vowels.
As for vowel duration, in line with previous findings, (Card, 1983; Zawaydeh, 1999),
64
short vowels show a larger F2 midpoint drop in emphatic positions than their long
counterparts. This also points to the gradient nature of emphasis spread.
F2 offset shows a significant drop in emphatic than in plain positions. This drop is
reported for all data and for words with initial and final target consonants as well. As
expected, a mirror effect is obtained here: F2 offset drop is larger in words with final
target consonants compared to words with initial consonants due to the fact that the
temporal interval between the emphatic segment and the vowel offset is shorter in
offset positions than in initial positions. High front vowels here show a larger F2 offset
drop than low front vowels, which, in turn, show a larger drop compared to high back
vowels. Evidently, emphatic consonants exert more influence on front vowels
compared to back vowels. This may be because back vowels already have lower F2 and
higher F1 values compared to front vowels. Therefore, there is less room for each
formant to change under emphasis compared to the room available for front vowels.
F3 is higher in emphatic than in plain positions. F3 onset is higher in emphatic
positions when all data are included in the measurements. This effect carries on to
words with initial target consonants but not to words with final target consonants. There
is an unexpected interaction with word-type where F3 onset rises by a larger degree in
words compared to nonwords. F3 midpoint is also higher in emphatic positions than in
plain ones for all data and also for words with initial target consonants. However, a
puzzling interaction with vowel quality is observed for F3 midpoint. While emphasis
65
causes F3 midpoint to rise, it does so for low front vowels and high back vowels. While
the difference is significant in low front vowels, it is not significant in high back vowels.
Surprisingly, emphasis causes F3 midpoint to decrease in high front vowels. F3 does
not rise at the midpoint position for words with final target consonants. In offset
positions, F3 rises in emphatic positions for all data, for words with final target
consonants, and shows a trend for words with initial target consonants. This effect is in
harmony with the findings on the directionality of emphasis spread within the
monosyllables.
While durations and formants show a host of significant distinctions in the acoustic
signal between plain and emphatic segments and environments, spectral moments fail
to show similar differences. The only significant finding that can be reported is the
interaction of manner with the center of gravity for the target consonants: while
emphatic stops have a higher center of gravity than their plain counterparts, fricatives
have a lower center of gravity compared to their plain counterparts. All other spectral
measurements on the target consonant and on the /b/ consonant as well fail to yield
significant differences.
As for the findings on locus equations, emphatic consonants have flatter slopes
compared to their plain counterparts. This finding it true for stops and fricatives and
also regardless of position of the target consonant. This suggests that emphatics are
more stable in terms of their resistance to effects of adjacent vowels. This is in line with
66
what Yeou (1997) found for the same set of consonants in Modern Standard Arabic
spoken by native speakers of the Moroccan dialect. He found that the same set of
emphatics emerged as a distinct class with the flattest slope values, which indicates
minimal coarticulation between the consonants and the vowels. Using data from
Cairene Arabic produced by three male speakers, Sussman, Hoemeke and Ahmed,
(1993) found that /d/ and /d¿/ had similar slopes but significantly differed in their yintercept values. They show that y-intercept was 1307 Hz for /d/ and 933 Hz for /d¿/.
The present study found no significant differences in slope values or in y-intercept
values for /d/ and /d¿/. While slope turns out to be a factor in defining emphatics as a
separate class of sounds, y-intercept fails to yield significant differences between plain
and emphatic consonants. Emphatics consistently had a higher y-intercept compared to
their plain counterparts, but this difference was not significant.
67
Chapter Three
Acoustic Study – the bisyllables
This chapter presents the acoustic findings on the bisyllables. In Chapter 2, analyses on
the monosyllables provided a first idea of emphasis spread by comparing words with
the target consonant in initial and final position. To get a better understanding of the
nature and domain/extent of spread, investigation of longer words will give us an idea
about emphasis spread to adjacent syllables. This chapter provides a first look at this
issue. Since spectral measures for the consonants in monosyllables did not really reveal
differences between plain and emphatic tokens, analysis of bisyllabic words will focus
on spectral measures of the vowels only. More precisely, this chapter aims to answer
the following questions:
1. Are the results obtained for monosyllables identical to those obtained for the syllable
which has the target consonant (the target syllable)?
2. Does emphasis spread beyond the target syllable?
3. Is emphasis spread attested for duration (for consonants and vowels) and/or vowel
formant frequency (F1, F2, F3)?
4. Does spreading to the right and left of the target syllable pattern similarly?
5. What is the effect of vowel quality on emphasis spread?
6. Is emphasis spread influenced by gender of the speaker?
68
Despite several limitations, see Section 1.7. above, previous research addresses some
of these questions. For example, Card (1983) found that emphasis, as an effect of F2
lowering, spreads beyond the target syllable unless in the presence of a ‘blocking
vowel,’ which she defines as one of the three long vowels /i, a, u/. Davis (1995) found
that in Palestinian Arabic, leftward emphasis was not blocked whereas rightward
emphasis was blocked by [+ high, - back] vowels. Based on these and similar previous
findings, see Section 1.4. above – and in accordance with the present findings on the
monosyllables, see Chapter Two, I can make the following hypotheses:
1. Emphatic consonants will be slightly longer than their plain counterparts only in
target syllables. Durational differences will disappear in adjacent syllables.
2. Vowel duration will not be affected by emphasis, neither in target syllables nor in
adjacent ones.
3.Vowel F1 and F3 values will rise, especially when measurements are taken closest to
the vowel in the same syllable with the emphatic target consonant. Vowel F2 will
decrease in the target vowel and in adjacent vowels as well
4. No gender interactions will be obtained.
5. No word/nonword differences will be obtained.
3. 1. Method
3.1.1. Subjects
Eight adult speakers (four males and four females) of UJA aged 16 to 26 years old
(average age was 20.6 years) participated in the study. Five of them (four males and
69
one female) were recruited from the Yarmouk University, Irbid, Jordan. The three other
females were recruited from the local community of Irbid: two of them were high
school students and one was a university graduate. All of the subjects spoke UJA. None
of them ever left Jordan. Four of the university students were majoring in religious
studies, one was majoring in psychology and one had a degree in English language and
literature. None of the subjects reported any known speech or hearing impairments.
They were paid $10 for their participation
3.1.2. Materials and Procedure
A long list of target words was prepared as part of a larger project. The wordlist in
question consisted of 80 target CV.CVC pairs (plain-emphatic). It included the four
emphatic consonants and their plain counterparts along with the six UJA vowels. It was
designed as follows: 24 pairs with the target consonant in word-initial position, and the
adjacent syllable was /-bat/ - yielding pairs such as /sabat/-/s¿abat/, /sibat/-/s¿ibat/,
/subat/-/s¿ubat/ etc.; another 24 pairs with the target consonant in a word-final position
and the adjacent syllable was /ta-/ - yielding pairs such as /tabas/-/tabas¿/, /tabis/-/tabis¿/,
/tabus/-/tabus¿/, etc.; and 32 pairs with the target consonant in word-medial position. In
this subset, all the target consonants were used, the filler consonant was /k/ and only
two UJA vowel pairs were used: long and short /a/ and /i/ - yielding pairs such as
/kasak/-/kas¿ak/, /kasik/-/kas¿ik/, /kasaak/-/kas¿aak/, etc. Thus, the study material
consisted of CV.bat, ta.bVC, and kV.CVk stimuli, see Appendix II. This design yielded
many nonwords that were all phonotactically acceptable in UJA. Subjects read the
70
target words in a carrier sentence: /'½iš.ki target word ka.'maan 'mar.rah/ [Say target
word again]. The wordlist included 12 pages, each having one column, and each
column consisted of 12 sentences – the first and last two sentences were fillers and did
not include target words. The best two of these three repetitions were used for analysis.
Choice was made based on the correct completion of the wordlist with the minimum
amount of mistakes. Since Arabic short vowels are not written in full letters, diacritics
were added to illustrate these vowels where they occurred. Speakers read the wordlist
three times. Recordings were carried out in a sound-proofed booth using a portable
solid-state recorder (Marantz PMD 671) and a unidirectional microphone (ElectroVoice
N/D767a). Each subject was recorded individually and each recording session took
approximately 70 minutes. Recordings were digitized at a 22050 Hz sampling rate.
Using Praat speech analysis software (Boersma and Weenink 2005), a script was
written to carry out automatic segmentation of the stimuli in order to isolate the target
words. Due to the relatively long duration of the recording sessions, some subjects
occasionally made mistakes in reading parts of the wordlist. In these instances, they
were instructed to repeat two lines before and after the line in which they made these
occasional mistakes. During the segmentation process through which the target sound
files were obtained, these lines which included possible mistakes were disposed of.
3.1.3. Measurements
The same method in Section 2.1.3 above was adopted here. However, slight changes
were applied due to the huge amount of data used in this section. The measurement
71
procedure this time was automatic. The cursors were placed by hand, using a waveform
and a spectrogram, at a number of locations in the speech signal. For stop consonants,
cursors were placed at the offset of the previous segment (the beginning of closure), at
the release of the burst, and at the onset of the following vowel. For fricatives, cursors
were placed at the onset and offset of the frication noise. For vowels, cursors were
placed at the start of F1 and the offset of F2. When the burst was very hard to locate, it
was estimated using auditory and visual cues. For the frequency values of the target
vowels, F1, F2 and F3 were measured at onset, midpoint and offset of the vowel. While
the cursors for the onset and offset points where easy to locate, the cursor for the vowel
midpoint was automatically derived. A script was written which took the measurements
for each segment and wrote them to an Excel file. Figure 1 below shows the placement
of the cursor at different points for an example target word: /sabat/
72
Figure 1: Sample measurements conducted for one stimulus word: /sabat/: (a) represents the
initial consonant; (b) the first vowel; (c) the closure of the second consonant; (d) the second
consonant burst; (e) the second vowel; (f) the closure of the third consonant; (g) the third
consonant burst; (h) the third consonant post-burst release.
For consonants in medial positions, the cursors where placed at the offset of the
previous vowel, where the three first formants ceased to line up, and at the onset of the
following vowel. For word-final consonants, the cursors were placed at the offset of the
previous vowel and at the beginning of the burst before the onset of the next stop
consonant of the carrier sentence. During this stage, different cues were utilized using
both the waveform and the spectrogram. Vowel formant frequency measurements were
based on LPC analysis using a 20-ms window.
73
3.2. Analysis
Different ANOVA tests were conducted on all measurements to analyze the data and
obtain the main findings. Multivariate analyses for different dependent variables were
also conducted to test interactions between the degree of emphasis and other
independent variables. Bonferroni post hoc tests were used for independent variables
with more than two levels, namely for vowel quality. Due to the fact that this chapter
deals with bisyllables, several dependent and independent variables were used, see
Table 1 below. The measurements mentioned in this table were made for each segment
in the word. Dependent variables are mentioned here in groups, that is; DV 1 refers to
durations of each consonant, and so on.
Table 1: Dependent and independent variables used in the study
No.
1
Dependent variables
Consonant duration
No.
1
2
Vowel duration
2
3
Word duration
Vowel F1 (onset, midpoint,
offset)
Vowel F2 (onset, midpoint,
offset)
Vowel F3 (onset, midpoint,
offset)
3
Independent variables
Consonant-type (plain, emphatic)
Vowel quality (high front, low front,
high back)
Adjacent vowel quality
4
Vowel length (long, short)
5
Adjacent vowel length
4
5
6
6
7
8
9
74
Target consonant position (initial,
medial, final)
Target consonant voicing (voiced,
voiceless)
Target consonant manner (stop,
fricative)
Speaker gender (male, female)
3.3. Results
In this section, I will report the main effects and the interactions for consonant and
vowel durations and for vowel formant frequencies. While all main effects and
interactions will be tested, I will focus on reporting those effects that show significant
differences between plain and emphatic environments.
3.3.1. Duration
Plain and emphatic target consonant durations were exactly equal (when rounded) at 93
ms, [F (1, 1782) = 0.118, p = 0.731]. A similar result was obtained for target
consonants in word-initial position where plain target consonants were 89 ms and
emphatic ones were 90 ms, [F (1, 374) = 0.036, p = 0.849]. The same also held for
target consonants in word-medial position where plain and emphatic target consonants
were 89 ms, [F (1, 1014) = 0.085, p = 0.771]. Word-final target consonant durations
were also not different. Plain target consonants in word-final position were 106 ms and
their emphatic counterparts were 111 ms, [F (1, 390) = 0.637, p = 0.425].
Except for one condition, other consonants in the stimuli did not show any significant
differences. That is to say, emphasis spread did not show on consonants in adjacent
positions. The only consonants which showed significant durational differences were
initial consonants in words where the target consonant was in word-final position. In
these stimuli, initial consonants in words with final plain targets were longer than their
counterparts in words with final emphatic targets, 83 ms and 78 ms, respectively. This
75
difference is significant, [F (1, 390) = 7.473, p = 0.007]. No interactions with target
consonant duration in word-initial positions were obtained.
As for vowels, measurements were taken in relation to the three possible target
consonant positions. Similar to target consonants, results on vowel durations did not
show significant differences. No significant durational differences were established for
vowels when the target consonant was in word-initial or word-final positions. However,
vowel durational differences were established when the target consonant was in word
medial position. Vowels in the same syllable with the emphatic consonant where the
target consonant was in word-medial position were significantly longer than their
counterparts in plain environments, 122 ms, and 114 ms, respectively. This difference
is significant: [F (1, 1014) = 4.779, p = 0.029]. There were no significant interactions
between target consonant manner, vowel length, and vowel quality.
3.3.2. Formant Frequency
Results on vowel formant frequencies are reported in this section in target and in
adjacent positions for F1, F2, and F3 at vowel onset, midpoint and offset. Main findings
will be reported first, then interactions between the degree of emphasis and other
independent variables. Where statistical analyses fail to show significant interactions,
results will not be reported.
76
3.3.2.1. F1
Since the target consonants appear in three different positions: word-initial, wordmedial and word-final, the results will be reported accordingly.
3.3.2.1.1. F1 target consonant word-initial
Under this condition, F1 for vowels in the same syllable with the target consonant, e.g.
[u] in [s¿ubat], was measured at three points in the vowel: onset, midpoint and offset.
Results show significant differences. In onset positions, F1 for vowels in plain
environments was 406 Hz, and 480 Hz for vowels in emphatic environments, [F (1,
374) = 52.364, p < 0.001]. The same significance was captured at vowel midpoint
positions, [F (1, 374) = 8.15, p =0.004], and at vowel offset positions, [F (1, 374) =
7.524, p = 0006]. Figure 2 shows these results:
77
600
508
480
Frequency (Hz)
500
*
*
465
474
454
*
406
Plain
400
Emphatic
300
200
Onset
Mid
Offset
Figure 2: F1 values in word-initial positions at onset, midpoint and offset positions for the
vowel in the same syllable with the target consonant.
A Consonant-type by Gender interaction shows that in emphatic environments, F1 at
onset position rises by 97 Hz for males and by 51 Hz for females. This difference is
significant,
[F (2, 372) = 5.19, p = 0.023]. Figure 3 shows these results.
78
600
Frequency (Hz)
500
Male
400
Female
300
200
Plain
Emphatic
Figure 3: Consonant-type by Gender interaction for F1 Onset
A similar result is obtained for vowel F1 at offset positions; a Consonant-type by
Gender interaction is obtained where vowel F1 in emphatic environments rises by 52
Hz for males and by 8 Hz for females. This difference is significant, [F (2, 372) = 4.42,
p = 0.036]; see Figure 4. No other interactions were obtained.
79
600
Frequency(Hz)
500
Male
Female
400
300
200
Plain
Emphatic
Figure 4: Consonant-type by Gender interaction for F1 Offset
Examining adjacency, e.g. [a] in [s¿ubat], F1 for the vowel in the right-adjacent syllable
was measured at three points: onset, midpoint and offset.
Except when F1 is measured in onset position, the differences shown here proved
insignificant. In onset position, F1 was significantly higher in emphatic environments,
[F (1, 374) = 15. 503, p < 0.001]. No significant results were obtained for F1
measurements for vowels in adjacent positions when these measurements were taken at
midpoint, [F (1, 374) = 1.245, p =0.265], or at offset, [F (1, 374) = 0.158, p =0.691].
Figure 5 plots these results:
80
700
611
Frequency (Hz)
600
*
620
561
542
539
532
Plain
500
Emphatic
400
300
Onset
Mid
Offset
Figure 5: F1 values in word-initial positions at onset, midpoint and offset positions for the
vowel right-adjacent to the syllable with the target consonant.
A significant Consonant-type by Gender interaction is obtained for the vowel in the
right-adjacent position when the measurement is taken at onset position: F1 in emphatic
environments rises by 29 Hz for males and only by 10 Hz for females. This difference
is significant, [F (2, 372) = 4.79, p = 0.029]; see Figure 6. No other significant
interactions were obtained.
81
600
Frequency(Hz)
500
Male
Female
400
300
200
Plain
Emphatic
Figure 6: Consonant-type by Gender interaction for F1 Onset for the vowel in the right-adjacent
position.
3.3.2.1.2. F1 target consonant word-medial
The stimuli used for this condition had only variations of two vowels: /i/ and /a/. For
vowels in word medial positions in the same syllable with the target consonant, e.g. [i]
in [kat¿ik], F1 was measured at three points in the vowel: onset, midpoint and offset.
The results show that F1 values at onset and mid positions were significantly higher in
emphatic environments compared to their plain counterparts, [F (1, 1014) = 68.897, p <
0.001], and [F (1, 1014) = 7.882, p = 0.005], respectively. F1 values at offset position
82
for the target vowel did not show significant differences, [F (1, 1014) = 2.300, p =
0.130]. Figure 7 below shows these results.
700
Frequency (Hz)
600
500
496
*
* 542
513
Plain
475
455
440
Emphatic
400
300
Onset
Mid
Offset
Figure 7: F1 values at onset, midpoint and offset positions for the vowel in the same syllable
with the word-medial target consonant.
No Consonant-type by Vowel Quality interactions were obtained for the target vowels
themselves, i.e. the vowels in the same syllable with the target consonant. However,
some interactions for the adjacent vowel were found. A Consonant-type by Vowel
Quality interaction was obtained for F1 Onset of the vowel to the left of the target one.
When the target vowel was /a/, F1 Onset for the left adjacent vowel rose by 71 Hz in
emphatic position compared to only 46 Hz when the target vowel was /i/. This
83
difference is significant, [F (1, 1011) = 8.05, p = 0.005]. Figure 8 below displays these
results.
800
Frequency (Hz)
700
600
/i/
500
/a/
400
300
200
Plain
Emphatic
Figure 8: Consonant-type by Vowel Quality interaction for F1 Onset of the vowel to the left
of the target one.
Another interaction, a Consonant-type by Adjacent Vowel Quality was obtained for F1
Onset, e.g. for /a/ in [kat¿ik]. Results show that in emphatic environments, F1 Onset
rises by (44 Hz) for /a/ and by 22 Hz for /i/. This difference is significant, [F (1, 1011)
= 17.77, p < 0.001]. Figure 9 below plots these results.
84
800
Frequency (Hz)
700
600
/i/
500
/a/
400
300
200
Plain
Emphatic
Figure 9: Consonant-type by Adjacent Vowel Quality interaction for F1 Onset of the vowel to
the left to the target one.
A similar result was obtained when the measurement was taken at the vowel midpoint.
A Consonant-type by Adjacent Vowel Quality was obtained for F1 Mid, e.g. for /a/ in
[kat¿ik]. Results show that in emphatic environments, F1 Mid rises by 73 Hz for /a/ and
by 43 Hz for /i/. This difference is significant, [F (1, 1011) = 8.49, p = 0.004]. Figure
10 below plots these results.
85
800
Frequency (Hz)
700
600
/i/
500
/a/
400
300
200
Plain
Emphatic
Figure 10: Consonant-type by Adjacent Vowel Quality interaction for F1 Mid of the vowel to
the left of the target one.
Similarly, a Consonant-type by Adjacent Vowel Quality interaction was obtained for
F1 Offset, e.g. for /a/ in [kat¿ik]. Results show that in emphatic environments, F1 Offset
rises by 82 Hz for /a/ and by 56 Hz for /i/. This difference is significant, [F (1, 1011) =
3.88, p = 0.049]. Figure 11 below plots these results.
86
800
Frequency (Hz)
700
600
/i/
500
/a/
400
300
200
Plain
Emphatic
Figure 11: Consonant-type by Adjacent Vowel Quality interaction for F1 Offset of the vowel
to the left of the target one.
Additionally, a Consonant-type by Gender interaction was also obtained for F1 Onset
of the vowel left to the target one. F1 Onset for males rises by 74 Hz in emphatic
environments compared to only 36 Hz for females. This difference is significant, [F (1,
1011) = 8.43, p = 0.004]. Figure 12 below displays this result.
87
600
Frequecy(Hz)
500
Male
400
Female
300
200
Plain
Emphatic
Figure 12: Consonant-type by Gender interaction for F1 Onset of the vowel to the left of the
target one.
When F1 measurements were taken at the middle of the left adjacent vowel, similar
results were obtained. A Consonant-type by Vowel Quality interaction was obtained for
F1 Mid of the vowel to the left of the target one. When the target vowel was /a/, F1 Mid
for the left adjacent vowel rose by 47 Hz in emphatic position compared to only 24 Hz
when the target vowel was /i/. This difference is significant, [F (1, 1011) = 3.97, p =
0.046]. Figure 13 below displays these results.
88
800
Frequency (Hz)
700
600
/i/
500
/a/
400
300
200
Plain
Emphatic
Figure 13: Consonant-type by Vowel Quality interaction for F1 Mid of the vowel to the left of
the target one.
Interestingly, when examining leftward adjacency, e.g. [a] in [kat¿ik], F1 shows a
significant rise when measurements were taken at onset, mid and offset positions. At
onset positions, F1 was significantly higher in emphatic environments compared to its
counterpart in plain environments, [F (1, 1014) = 47.631, p < 0.001]. The same applied
to F1 values when the measurement was taken at midpoint, [F (1, 1014) = 43.878, p <
0.001] and at offset, [F (1, 1014) = 87.760, p < 0.001]. Figure 14 shows these results.
89
700
Frequency (Hz)
600
513
400
485
*
451
500
*
384
Plain
Emphatic
*
418
414
300
Onset
Mid
Offset
Figure 14: F1 values in word-medial positions at onset, midpoint and offset positions for the
vowel left-adjacent to the syllable with the target consonant.
3.3.2.1.3. target consonant word-final
For vowels in word final positions in the same syllable with the target consonant, e.g. [i]
in [tabis¿], F1 was measured at three points in the vowel: onset, midpoint and offset.
Results show that target vowel F1 was significantly higher in emphatic environment
compared to its plain counterpart at two positions: onset and offset of the vowel.
Differences obtained when the measurements were taken at vowel midpoint were not
significant. For onset positions, [F (1, 389) = 5.137, p = 0.024]; for mid vowel positions,
[F (1, 389) = 2.975, p = 0.085]; and for vowel offset, [F (1, 389) = 20.402, p < 0.001].
Figure 15 below shows these results.
90
700
Frequency (Hz)
600
537
*
500
513
505
490
*
468
451
Plain
Emphatic
400
300
Onset
Mid
Offset
Figure 15: F1 values in word-final positions at onset, midpoint and offset positions for the
vowel in the same syllable with the target consonant.
F1 for the vowel in the left adjacent positions, e. g. [a] in [tabis¿], was significantly
higher in emphatic environments compared to plain ones when measurements were
taken at three positions: onset, [F (1, 389) = 72.503, p = 0.00], vowel midpoint, [F (1,
389) = 40.933, p = 0.00], and vowel offset, [F (1, 389) = 44.162, p = 0.00]. Figure 16
below shows these results.
91
700
Frequency (Hz)
600
*
540
*
500
580
*
522
534
497
485
Plain
Emphatic
400
300
Onset
Mid
Offset
Figure 16: F1 values in word-final positions at onset, midpoint and offset positions for the
vowel left-adjacent to the syllable with the target consonant.
A significant Consonant-type by Gender interaction was obtained for F1 of the left
adjacent vowel when the measurement was taken at the offset. F1 in emphatic
environments rises for males by 79 Hz compared to females 30 Hz. This difference is
significant, [F (2, 387) = 4.17, p = 0.042]; see Figure 17 below.
92
600
Frequecy(Hz)
500
Male
400
Female
300
200
Plain
Emphatic
Figure 17: Consonant-type by Gender interaction for F1 Offset for the vowel in the left adjacent
position.
3.3.2.2. F2
Since the target consonants appear in three different positions: word-initial,
word-medial and word-final, the results will be reported accordingly.
3.3.2.2.1. target consonant word-initial
For vowels in the same position with the target consonant, e.g. [u] in [s¿ubat], F2 was
measured at three points: onset, midpoint and offset. Where target consonants appear in
word-initial positions, the results show that F2 values for vowels in emphatic
environments drop significantly. However, the drop is only significant when F2 is
93
measured at onset position. F2 in vowel onset position is significantly lower in
emphatic environments compared to plain ones, [F (1, 374) = 32.261, p <0.001]. When
measured at vowel midpoint, differences were not significant, [F (1, 374) = 2.789, p =
0.096]; and similarly for vowel offset, [F (1, 374) = 0.862, p = 0.354]. Figure 18 below
shows the results:
2000
1983
1875
*
1800
1767
1701
1702
Frequency (Hz)
1646
1600
Plain
Emphatic
1400
1200
1000
Onset
Mid
Offset
Figure 18: F2 values in word-initial positions at onset, midpoint and offset positions for the
vowel in the same syllable with the target consonant.
A Consonant-type by Vowel Quality interaction was obtained for F2 in word-initial
positions. In emphatic environments, F2 for the vowel in word-initial target positions
dropped by 437 Hz for /i/, 421 Hz for /a/ and only by 32 Hz for /u/. This difference is
significant, [F(2, 370) = 14.21, p<0.001]. Figure 19 shows these results:
94
2600
2400
Frequency (Hz)
2200
2000
Plain
1800
Emphatic
1600
1400
1200
1000
/i/
/a/
/u/
Figure 19: Consonant-type by Vowel Quality interaction for F2 Onset in target positions.
A Consonant-type by Gender interaction is also obtained for F2 in word-initial
positions. In emphatic environments, F2 drops by 458 Hz for females and by 105 Hz
for males, [F (2, 372) =13.31, p<0.001], Figure 20 below shows these results:
95
2600
2400
Frequency(Hz)
2200
2000
Plain
1800
Emphatic
1600
1400
1200
1000
Male
Female
Figure 20: Consonant-type by Gender interaction for F2 Onset in target positions.
Results show that F2 for vowels in right adjacent positions, e.g., [a] in [s¿ubat], was
significantly lower in emphatic environments compared to plain ones. In onset position,
the F2 drop for vowels in emphatic positions was significant, [F (1, 374) = 32.512, p <
0.001]. A similar result was obtained when the measurement was taken at midpoint, [F
(1, 374) = 50.212, p < 0.001], and when the measurement was taken at vowel offset, [F
(1, 374) = 13.453, p < 0.001]. Figure 21 plots these results:
96
2000
1800
1749
Frequency (Hz)
1641
1600
*
1522
1496
1400
*
1670
Plain
*
Emphatic
1371
1200
1000
Onset
Mid
Offset
Figure 21: F2 values at onset, midpoint and offset positions for the vowel right-adjacent to the
syllable with the word-initial target consonant.
3.3.2.2.2. target consonant word-medial
For vowels in word medial positions in the same syllable with the target consonant, e.g.
[i] in [kat¿ik], results show that F2 drops significantly in emphatic positions when
measurements were taken at vowel onset, [F (1, 1014) = 340.498, p < 0.001]; vowel
midpoint, [F (1, 1014) = 104.776, p < 0.001]; and vowel offset, [F (1, 1014) = 70.116, p
< 0.001]. Figure 22 below shows these results.
97
2400
2267
2200
*
2058
2103
Frequency (Hz)
2052
2000
*
Plain
1823
1800
Emphatic
*
1626
1600
1400
Onset
Mid
Offset
Figure 22: F2 values at onset, midpoint and offset positions for the vowel in the same syllable
with the word-medial target consonant.
A Consonant-type by Vowel Quality interaction is obtained where F2 in emphatic
environments drops by 349 Hz for /a/ and only by 92 Hz for /i/, [F(1, 1011) =65.42,
p<0.001]. Figure 23 plots these results:
98
2600
2400
Frequency(Hz)
2200
2000
Plain
1800
Emphatic
1600
1400
1200
1000
/i/
/a/
Figure 23: Consonant-type by Vowel Quality interaction for F2 Mid in target positions.
A Consonant-type by Gender interaction is also obtained where F2 for females drops in
emphatic environments by 264 Hz, and by 156 Hz for males. This is a significant
difference, [F(1, 1011) =4.74 , p=0.030]. Figure 24 shows the results:
99
2600
2400
Frequecy(Hz)
2200
2000
Male
1800
Female
1600
1400
1200
1000
Plain
Emphatic
Figure 24: Consonant-type by Gender interaction for F2 Mid in target positions.
A Consonant-type by Adjacent Vowel Quality interaction is also obtained for F2 in
word-medial positions for the vowels adjacent to the target one where F2 drops by 408
Hz for the vowel /a/ and by 113 Hz for the vowel /i/. This is a significant difference,
[F(1, 1011) =92.94, p<0.001]. Figure 25 below shows the results:
100
2600
2400
Frequency (Hz)
2200
2000
Plain
1800
Emphatic
1600
1400
1200
1000
/i/
/a/
Figure 25: Consonant-type by Adjacent Vowel Quality interaction for F2 Onset in adjacent
positions.
In adjacent positions, a Consonant-type by Gender interaction is also obtained where F2
in emphatic environments drops by 323 Hz for females and by 204 Hz for males. This
difference proves significant, [F (1, 1011) =11.00 , p=0.001]. See Figure 26.
101
2600
2400
Frequecy(Hz)
2200
2000
Male
1800
Female
1600
1400
1200
1000
Plain
Emphatic
Figure 26: Consonant-type by Gender interaction for F2 Onset in adjacent positions.
For vowels in left adjacent positions, e.g., [a] in [kat¿ik], F2 drops significantly in
emphatic positions when measurements were taken at vowel onset, [F (1, 1014) =
166.839, p < 0.001]; vowel midpoint, [F (1, 1014) = 207.395, p < 0.001]; and vowel
offset, [F (1, 1014) = 464.616, p < 0.001]. Figure 27 below shows these results.
102
2400
2385
2231
Frequency (Hz)
2200
*
2121
2127
*
2000
Plain
1880
Emphatic
1800
*
1662
1600
1400
Onset
Mid
Offset
Figure 27: F2 values at onset, midpoint and offset positions for the vowel left-adjacent to the
syllable with the word-medial target consonant.
3.3.2.2.3. target consonant word-final
For vowels in word final positions in the same syllable with the target consonant, e.g. [i]
in [tabis¿], F2 was measured at three points in the vowel: onset, midpoint and offset. At
all measurement points, F2 drops significantly in emphatic environments compared to
the plain counterparts. At onset positions, [F (1, 389) = 7.389, p = 0.007]; at mid
positions, [F (1, 389) = 8.105, p = 0.005]; and at offset positions, [F (1, 389) = 29.169,
p < 0.001]. Figure 28 shows these results.
103
2000
1822
1800
1736
Frequency (Hz)
1666
1600
*
*
1508
1559
*
1595
Plain
Emphatic
1400
1200
1000
Onset
Mid
Offset
Figure 28: F2 values in word-final positions at onset, midpoint and offset positions for the
vowel in the same syllable with the target consonant.
F2 for the vowel in the left-adjacent positions, e.g. [a] in [tabis¿], was significantly
lower in emphatic environments compared to plain ones. At onset positions, [F (1, 389)
= 100.231, p < 0.001]; at mid positions, [F (1, 389) = 84.499, p < 0.001]; and at offset
positions, [F (1, 389) = 55.099, p < 0.001]. Figure 29 shows these results.
104
2000
1883
*
1717
Frequency (Hz)
1800
1811
*
1622
1600
1652
*
1452
Plain
Emphatic
1400
1200
1000
Onset
Mid
Offset
Figure 29: F2 values at onset, midpoint and offset positions for the vowel left-adjacent to the
syllable with the word-final target consonant.
3.3.2.3. F3
Since the target consonants appear in three different positions: word-initial,
word-medial and word-final, the results will be reported accordingly.
3.3.2.3.1. target consonant word-initial
For vowels in the same position with the target consonant, e.g. [u] in [s¿ubat], F3 was
measured at three points: onset, midpoint and offset. The results show that F3 is higher
for vowels in emphatic environments. This increase is significant regardless of the
position in the vowel at which the measurement was taken. In onset position, F3 rises
105
significantly in emphatic environments, [F (1, 374) = 18.819, p < 0.001]. When the
measurement is taken at vowel midpoint, a similar result is revealed, [F (1, 374) =
4.732, p = 0.030]. And the same is true for measurements taken at vowel offset
positions, [F (1, 374) = 16.143, p < 0.001]. Figure 30 plots these results:
3200
*
3000
3094
2990
* 3087
3027
2991
*
Frequency (Hz)
2883
2800
Plain
2600
Emphatic
2400
2200
2000
Onset
Mid
Offset
Figure 30: F3 values in word-initial positions at onset, midpoint and offset positions for the
vowel in the same syllable with the target consonant.
For F3 measurements for vowels in right adjacent positions, e.g. [a] in [s¿ubat], the
results show that F3 was significantly higher in emphatic environments compared to
plain ones only when the measurements were taken at onset and midpoint positions, not
at offset positions.
106
Vowel F3 in onset position is significantly higher in emphatic compared to plain
environments, [F (1, 374) = 13.328, p < 0.001]. A similar result is obtained for vowel
F3 at midpoint, [F (1, 374) = 8.376, p = 0.004]. However, no significant difference was
obtained for vowel F3 in offset position, [F (1, 374) = 2.972, p = 0.086]. Figure 31 plots
these results:
3200
2956
3000
Frequency (Hz)
*
2800
*
2723
2803
3000
2857
2799
Plain
2600
Emphatic
2400
2200
2000
Onset
Mid
Offset
Figure 31: F3 values at onset, midpoint and offset positions for the vowel right adjacent to the
syllable with the target consonant.
3.3.2.3.2. target consonant word-medial
For vowels in word medial positions in the same syllable with the target consonant, e.g.
[i] in [kat¿ik], results show that F3 rises significantly in emphatic environments when
measurements are taken at onset and offset positions but not at vowel midpoint. At
107
onset positions, [F (1, 1014) = 32.793, p < 0.001]; at vowel midpoint, [F (1, 1014) =
2.400, p = 0.122]; and at vowel offset, [F (1, 1014) = 4.267, p = 0.039]. Figure 32
below plots these results.
3200
Frequency (Hz)
3000
* 3008
2927
2932 2956
* 2975
2940
2800
Plain
2600
Emphatic
2400
2200
2000
Onset
Mid
Offset
Figure 32: F3 values at onset, midpoint and offset positions for the vowel in the same syllable
with the word-medial target consonant.
Surprisingly, for vowels in left adjacent positions, e.g., [a] in [kat¿ik], F3 drops
significantly in emphatic environments when measurements were taken at onset and
midpoint positions, but expectedly rises when measurements were taken at offset
positions. F3 onset and midpoint for the left adjacent vowel were significantly lower in
emphatic environments, [F (1, 1014) = 27.801, p < 0.001], and [F (1, 1014) = 6.439, p
= 0.011], respectively. F3 offset for the left adjacent vowel was significantly higher in
108
emphatic environments, [F (1, 1014) = 21.142, p < 0.001]. Figure 33 below shows
these results.
3200
3028
Frequency (Hz)
3000
*
2944
2952
*
2914
* 2998
2932
2800
Plain
2600
Emphatic
2400
2200
2000
Onset
Mid
Offset
Figure 33: F3 values at onset, midpoint and offset positions for the vowel left-adjacent to the
syllable with the target consonant.
3.3.2.3.3. target consonant word-final
For vowels in word final positions in the same syllable with the target consonant, e.g. [i]
in [tabis¿], F3 was measured at three points in the vowel: onset, midpoint and offset. At
all measurement points, vowel F3 significantly rises in emphatic environments
compared to plain counterparts. F3 is higher in emphatic environments when
measurements were taken at onset positions, [F (1, 389) = 15.857, p = 0.00]; at
109
midpoint, [F (1, 389) = 12.830, p = 0.00]; and at offset, [F (1, 389) = 56.965, p = 0.00].
Figure 34 below displays these results.
3139
3200
*
3000
2972
*
3074
2975
*
2958
Frequency (Hz)
2858
2800
Plain
2600
Emphatic
2400
2200
2000
Onset
Mid
Offset
Figure 34: F3 values at onset, midpoint and offset positions for the vowel in the same syllable
with the word-final target consonant.
F3 for the vowel in the left-adjacent positions, e.g. [a] in [tabis¿], was significantly
higher in emphatic environments compared to plain ones. However, this result was
obtained when measurements were taken at vowel midpoint and offset positions but not
at vowel onset positions. At onset positions, vowel F3 did not show significant
differences, [F (1, 389) = 3.249, p = 0.072]. At vowel midpoint, F3 was significantly
higher in emphatic environments, [F (1, 389) = 6.061, p = 0.014], and similar results
were obtained when measurements were conducted at vowel offset positions, [F (1, 389)
= 14.876, p = 0.00]. Figure 35 below shows these results.
110
3200
Frequency (Hz)
3000
2891
2928
*
2839
2892
2800
2849
*
2760
Plain
2600
Emphatic
2400
2200
2000
Onset
Mid
Offset
Figure 35: F3 values at onset, midpoint and offset positions for the vowel left-adjacent to the
syllable with the target consonant.
3.4. Results Summary
Tables 2, 3, and 4 below summarize the main results on the bisyllables. Since segment
durations did not show any consistent effects, we will focus on spectral measures here
and in the subsequent discussion. This arrow (y) represents an increase in formant
frequency in emphatic position, the other one (z) represents a decrease in formant
frequency in emphatic position. The numbers in each table under Difference show the
magnitude of the increase or decrease (in Hz).Wherever numbers are shown in
parentheses, the difference was not significant.
111
Table 3: Emphasis in words with initial target consonants.
sa.bat
F1 onset
F1 mid
F1 offset
F2 onset
F2 mid
F2 offset
F3 onset
F3 mid
F3 offset
Vowel 1 Difference (Hz)
74
y
43
y
20
y
282
z
(108)
z
(56)
z
104
y
60
y
108
y
Vowel 2
y
y
z
z
z
z
y
y
y
Difference
19
(9)
(7)
125
119
79
80
58
(44)
Vowel 2
F1 onset
y
F1 mid
y
F1 offset
y
F2 onset
z
F2 mid
z
F2 offset
z
F3 onset
y
F3 mid
y
F3 offset
y
Difference
56
29
(20)
426
280
209
81
(24)
35
Vowel 1
F1 onset
y
F1 mid
y
F1 offset
y
F2 onset
z
F2 mid
z
F2 offset
z
F3 onset
y
F3 mid
y
F3 offset
y
Difference
37
40
37
166
189
200
(37)
53
89
F1 onset
F1 mid
F1 offset
F2 onset
F2 mid
F2 offset
F3 onset
F3 mid
F3 offset
Table 4: Emphasis in words with medial target consonants.
ka.sak
F1 onset
F1 mid
F1 offset
F2 onset
F2 mid
F2 offset
F3 onset
F3 mid
F3 offset
Vowel 1
y
y
y
z
z
z
z
z
y
Difference (Hz)
34
62
71
264
351
465
84
38
66
Table 5: Emphasis in words with final target consonants.
ta.bas
F1 onset
F1 mid
F1 offset
F2 onset
F2 mid
F2 offset
F3 onset
F3 mid
F3 offset
Vowel 2
y
y
y
z
z
z
y
y
y
Difference (Hz)
22
(24)
54
158
177
227
114
99
181
112
3.5. Discussion
In line with previous studies and similar to the present results for the monosyllables, the
main findings indicate a significant increase for vowel F1 and F3 values in emphatic
target positions and a significant decrease for vowel F2. However, results are more
complex for bisyllables. In words with initial target consonants, e.g. [s¿a.bat], F1
significantly rises for the target vowel in all measurement positions but this difference
fades away in the right-adjacent vowel when measurements are taken at midpoint and
offset positions. For words with a medial target consonant, e.g., [ka.s¿ak], F1 values for
the vowel in the same syllable with the target consonant rises at onset and midpoint but
no significant differences were obtained for target vowel offset. Interestingly, for the
vowel in the left-adjacent position, i.e. crossing a syllable, F1 rises at the three
measurement points. As for words with target consonants in final positions, i.e.
[ta.bas¿], F1 rises at all three measurement points for the target vowel and continues to
rise for F1 onset and offset positions, not for F1 midpoint. These results suggest that
for the most part, F1 increases in emphatic positions when the vowel is in the same
syllable with the target consonant. The same effect seems to spread leftwards more
easily than rightward. Card (1983) mentions leftward emphasis spread – as observed
through F2 lowering – being more attested compared to rightward emphasis. However,
her results are based on monosyllables. In the present study, leftward emphasis spread
crosses the syllable boundary compared to rightward emphasis spread, which seems to
be confined to the target syllable.
113
F2 shows a significant drop in emphatic environments. In words with initial target
consonants, F2 drops significantly at vowel onset but the drop is insignificant at vowel
midpoint and offset. The drop is attested in vowels right-adjacent to the target
consonant at all measurement points. This only occurs in words with initial target
consonants. In words with medial and final target consonants, F2 drops significantly in
emphatic positions at the three measurement points and for the two vowels: the target
and the left-adjacent ones. In accordance with our previous observations, F2 drops in
the target positions and left of the syllable with the target vowel. This gives more
evidence that leftward emphasis spread is more prominent in UJA.
F3 shows a significant increase in emphatic environments. In words with initial targets,
vowels in the same syllable show a significant increase. This effect spreads to the
vowel in the right-adjacent syllable when measurements are taken at vowel onset and
midpoint, not at vowel offset. In words with medial target consonants, F3 rises in target
positions when the measurements are taken at vowel onset and offset but not at vowel
midpoint. For the vowel left of the target syllable, F3 surprisingly drops at vowel onset
and midpoint but rises significantly at vowel offset. In words with final target
consonants, F3 rises in the vowel in the same syllable with the target consonant when
the measurements are taken at vowel midpoint and offset, but not at vowel onset.
However, F3 significantly increases in left-adjacent vowels at the three measurement
points.
Despite only a few odd findings in the results, mentioned above, there is a clear effect
of emphasis on vowels whereby F1 and F3 increase while F2 drops in emphatic
114
positions; and where leftward emphasis spread – as an effect of F1/F3 increase and F2
drop – is more obvious compared to rightward emphasis spread such that the former
spans across the syllable boundary. This shows that anticipatory coarticulation is more
common than preservatory coarticulation. Emphasis spread is asymmetrical in nature
and "pharyngealization (like velarization) is anchored on the onset phase of the primary
articulation and for that reason tends to spread in an anticipatory manner, affecting the
formants of preceding segment(s) more than the formants of following segment(s),"
Watson (1999: 293). See also Laver 1994; Davis 1995; Ladefoged and Maddieson 1996
for similar discussions on the prevalence of anticipatory coarticulation. Watson,
supported by Ladefoged and Maddieson (1996), ascribes this asymmetrical nature of
emphasis spread to secondary articulation, (1999: 298). She explains that in emphasis,
"the pharynx narrows prior to the hold phase of the primary articulation; thus,
pharyngealization is anchored more on the onset of the primary articulation, which
results in the anticipatory nature of spread of pharyngealization".
The analyses in this chapter give room to a wider range of interactions that can enable
us to better understand the acoustic correlates of emphasis. A number of independent
variables were tested, see Table 1 above. Among the different variables that were
examined, two seem to influence the degree of emphasis and/or emphasis spread:
gender of the speaker and vowel quality. As an effect of emphasis, F1 values increase
more in males than in females whereas F2 drops significantly more in females
compared to males; see Figures 3, 4, 6, 12, and 17 above.
115
With regard to vowel quality, it is rather obvious that the vowels that are transparent to
emphasis, i.e., do not seem to block emphasis spread are /a/ and /i/, where results show
that sometimes emphasis as an effect of F1 increase is significantly more evident in /a/
compared to /i/. However, the high back vowel /u/ seems to block emphasis spread.
This result holds for both the target vowels and the vowels in adjacent positions. Since
emphasis involves a clear secondary back articulation, it is not surprising that it affects
front and low vowels more than back vowels. This is what the data on bisyllables
confirm.
The acoustic findings on the bisyllables substantiate earlier findings on the
monosyllables and give a better idea of emphasis spread and its extent/domain. It is fair
to claim that most effects of emphasis can be observed in the vowels within the target
or adjacent syllables – despite a general feeling among native speakers of UJA, and
MSA as well, that the target consonant triggers the emphatic perception of the
consonant. These acoustic findings will be tested in the next chapter to verify, from the
point of view of the listener, whether the vowels or the consonants carry the effects of
emphasis.
116
Chapter Four
The perception study
The main findings of the acoustic study for the monosyllables showed that emphatic
stops in word-final positions were consistently longer than their plain counterparts.
While this was the only significant difference in terms of duration, the frequency
measurements on the vowels provided deeper and more vivid insights into the nature of
emphasis. Vowel F1 and F3 rose or tended to rise while F2 was consistently lower in
emphatic positions compared to their plain counterparts. However, different
interactions showed that some independent variables affected the degrees of increase or
decrease of formant frequency such as vowel quality, position of the target consonant
and vowel length. In the present chapter, I investigate whether some of the findings of
the acoustic study map on to perception. Namely, whether the difference in formant
frequencies between plain and emphatic obstruents and its variability due to vowel
quality affect perception. Of course, investigating every aspect of the difference
between plain and emphatic consonants and vowels in their neighborhood requires
extensive efforts and experimental designs to account for every interaction between the
consonant-type (plain and emphatic) and all other independent variables. This is
beyond the scope of this chapter. For the purposes of this study, I included two
independent variables: manner of articulation: a stop /t-t¿/ and a fricative /s-s¿/, and
vowel quality: low front /a/, and high back /u/.
117
While high front vowels /i-i:/ and low front vowels /a-a:/ maintain similar effects on
emphasis, cf. Section 2.3.2.2., high back vowels /u-u:/ show different effects. For low
front vowels in word-initial position, F2 onset drops by 529 Hz in emphatic positions
compared to their plain counterparts, whereas F2 onset for high back vowels drops only
by 375 Hz. F2 midpoint drops by 412 Hz for low front vowels and only by 184 Hz for
high back vowels, both in word-initial position. And finally, F2 offset drops by 280 Hz
for low front vowels and only by 42 Hz for high back vowels in the same position. F3
midpoint rises in emphatic positions for low front vowels but it does not show
significant differences for high back vowels. No other interactions were found for /a-a:/
or /u-u:/ with consonant-type that affected F1 or F3 measurements. All measurements
on duration and spectral moments for /b/ in plain and emphatic positions yielded no
differences.
Given these results, it is crucial to test the effects of the target consonants and target
vowels (that is, those vowels in the same syllable with the target consonant) on the
perception of emphasis. To investigate this issue, the present experimental design
utilizes splicing the segments that make up the target word and rearranging them such
that the one segment that is crucial to emphasis can be singled out. The first
requirement for this design is to establish whether subjects detect emphasis in spliced
target words. That is, to examine whether subjects can detect the plain or emphatic
target consonants in words that have been spliced and reassembled. This provides the
118
grounds for arguing that splicing per se does not affect perception. The following are
the questions this experiment attempts to address:
1. Is the perception of emphasis primarily driven by the target consonant or by the
target vowel?
2. Is the perception of emphasis affected by the manner of articulation of the emphatic
consonant?
3. Is the perception of emphasis affected by vowel quality?
To answer these questions, an experimental design is adopted to find out what
segment is crucial to the perception of emphasis and under what conditions, the
reaction times to the different conditions, and the rating of the subjects with regard to
their confidence level in their answers. Based on the findings of the acoustic study on
monosyllables and aided by the literature on emphasis, I make the following
hypotheses:
1. The perception of emphasis is not affected by the presence or absence of the
emphatic target consonant.
2. The perception of emphasis is governed by the target vowels.
3. Vowel quality affects the perception of emphasis: emphasis is more clearly
perceivable in low front vowels than in high back vowels.
119
4. 1. Method
4.1.1. Subjects
20 participants, 16 males and 4 females, whose native dialect was UJA took part in the
experiment. 16 of them were recruited from Kansas (Kansas City, Lawrence and
Manhattan) and 4 were recruited from Arkansas (Fayetteville). All of them were
college-age students, employees or their spouses. None of them reported having any
hearing or speech impairments.
4.1.2. Materials and Procedure
Four pairs of plain and emphatic words produced by the same subject were selected
from the sound files used in the acoustic study on the monosyllables: /tab-t¿ab/, /tubt¿ub/ /sab-s¿ab/, and /sub-s¿ub/. Table 1 shows the duration and formant frequency
measurements for each of these sounds.
120
Table 1: The duration and formant frequency measurements for the perception stimuli
Word
TC duration
Vowel duration
/b/ duration
V_F1 onset
V_F1 midpoint
V_F1 offset
V_F2 onset
V_F2 midpoint
V_F2 offset
V_F3 onset
V_F3 midpoint
V_F3 offset
sab
s¿ab
sub
s¿ub
tab
t¿ab
tub
t¿ub
161
145
126
127
114
149
151
139
117
134
120
107
109
126
136
100
105
90
106
126
126
108
126
109
368
443
354
374
372
477
379
434
473
505
404
434
480
525
416
411
390
301
309
328
345
385
214
345
1457
1117
1387
1308
1568
1248
1425
1833
1432
1157
1230
1527
1455
1146
1232
1037
1252
1102
1029
1140
1316
1026
936
1363
2464
2405
2551
2920
2714
2649
2745
3105
2385
2538
2378
3262
2377
2579
2447
2486
2246
2352
2291
2390
3113
2281
2466
3277
Each of these words was cut into its three segments, and then spliced back together
such that every word that the subjects heard had been spliced. This condition ensures
that none of the words used in the experiment is presented intact. Each segment was
defined in accordance with the procedure used in the acoustic study on the
monosyllables. Specifically, the same criteria adopted for the duration measurements of
the stops, the fricatives and the vowels in Chapter 2 were also adopted for these stimuli,
cf. Section 2.1.3. Two other conditions were applied to each sound file. For each pair,
the plain and the emphatic target consonants were cut and swapped. This created two
stimuli: the plain word with the emphatic target consonant, and the emphatic word with
the plain target consonant. Then, the same process is repeated for the vowels: for each
pair, the vowels in the plain and the emphatic environments were cut and swapped,
yielding two other stimuli: the plain word with the vowel from the emphatic word and
the emphatic word with the vowel from the plain word. The same procedure was used
121
for the four pairs of target words. Every condition is applied to a copy of the original
‘intact’ target word. Each condition was repeated 10 times for a total of 240 sound files.
Table 2 below shows the conditions used to create the stimuli. Letters in upper case
represent the segments that are emphatic; letters in lower case represent the segments
that are plain.
Table 2: The conditions used in the perception study
No.
1
2
3
4
5
6
7
8
9
10
11
12
Condition
original words spliced
change target Stop
change target Vowel
original words spliced
change target Stop
change target Vowel
original words spliced
change target Fricative
change target Vowel
original words spliced
change target Fricative
change target Vowel
Plain
tab
T¿ab
tAb
tub
T¿ub
tUb
sab
S¿ab
sAb
sub
S¿ub
sUb
Emphatic
T¿AB
tAB
T¿aB
T¿UB
tUB
T¿uB
S¿AB
sAB
S¿aB
S¿UB
sUB
S¿uB
The experiment was conducted using the Superlab software (Version 2). A picture file
containing the text “How confident are you of your answer” written in Modern
Standard Arabic orthography was created in Microsoft Paint. The sound files were used
to determine what target consonant the subjects heard and also to collect reaction time
data; the picture files were used to measure confidence ratings for each sound file.
Subjects heard the sound files on a Sony MDR-7502 Dynamic Stereo Headphone and
used a laptop with defined keyboard buttons to record their responses.
122
Each subject took a practice session of 24 trials, followed by the experimental set of
240 trials. No feedback was given during the practice session. Each trial consisted of a
sound file and a picture file. Trials were randomized by Superlab, and a 1000 ms ISI
separated each event. Subjects were tested individually and were instructed that they
would hear a list of monosyllabic words starting with one of the target consonants: /t, t¿,
s, s¿/. They were asked to push a response button that represented the initial consonant
of the word that they heard. Reaction time was measured from the onset of the sound
file until the subject pushed the response button. Once they responded to the sound file,
a picture file would appear asking them how confident they were of their answers. They
were asked to push a response button indicating their confidence level at a scale from 1
(least confident), to 5 (most confident). Therefore, every time the subjects heard a
sound file, they responded to it, and then the picture file with the question “How
confident are you of your answer” followed. Average time for completing the
experiment was around 16 minutes per subject.
4.2. Analysis
For each subject, responses to the sound file, reaction time and confidence ratings were
collected. These were the dependent variables. For the sound file responses and the
confidence ratings, a simple frequency count was used. Different ANOVA’s were used
to compare reaction times among the different conditions. The results will be presented
for stops first, then for fricatives. Though it was expected that subjects would respond
with a plain or an emphatic segment of the same manner, some subjects responded with
123
a different manner. In other words, where a certain condition was applied to /t/ or /t¿ /,
responses were expected to be confined to these two stops. However, in some of these
conditions, a few subjects responded with an /s/ or /s¿ /.
4.3. Results
In this section, I report the responses to the three conditions applied to the sound files,
the reaction times, and the confidence ratings.
4.3.1. Original words
Under this condition, the original sound files were cut into their segments and spliced
back together. Subjects detected the correct target consonants at rates ranging from
99% correct responses for /T¿UB/ and /S¿UB/ to 100% correct responses for /tab/, /sab/,
/S¿AB/ and /sub/. /T¿AB/ and /tub/ fell in the middle with 99.5% correct responses.
Despite the fact that the results go in the expected direction for this condition, the
reaction time data in each case tell a slightly different story. A one-way ANOVA on
reaction times shows that subjects responded significantly faster to /t/ in /tab/ (992 ms)
compared to /t¿/ in /T¿AB/ (1151 ms), [F (1, 398) = 13.639, p < 0.001]. This difference
is not detected for stops in words with high back vowels. Subjects detected /t/ in /tub/
with a mean RT of 1058 ms, and /t¿/ in /T¿UB/ with a mean RT of 1123 ms. This is not
a significant difference, [F (1, 398) = 2.835, p = 0.092]. While subjects responded faster
to /t/ than /t¿/ in the low front vowel context, mean RT for /s/ in /sab/ is 1059 ms and
124
986 ms for /s¿/ in /S¿AB/. This difference is significant, [F (1, 398) = 5.188, p = 0.023].
No significant difference is obtained in RT for /s/ in /sub/ (1098 ms) and /s¿/ in /S¿UB/
(1124 ms), [F (1, 398) = 0.389, p = 0.532].
The confidence rating data show that subjects were highly confident in most of their
answers. Table 3 shows the mean confidence for each condition.
Table 3: Mean confidence in responses for spliced words
Condition
tab
T¿AB
tub
T¿UB
sab
S¿AB
sub
S¿UB
total
Confidence rating
4.9
4.7
4.8
4.7
4.9
4.9
4.9
4.7
4.8
4.3.1.1. Summary
The results for the first condition show that subjects detected the target consonants with
a high degree of accuracy ranging from 99 – 100%. Reaction time data show that
subjects detected plain stops faster than emphatic stops in the environment of low front
vowels and they detected emphatic fricatives faster than plain fricatives in the same
vocalic environment. In the environment of high back vowels, no significant
differences were found. And finally, subjects expressed a high confidence rating of
their answers ranging from 4.7 to 4.9.
125
4.3.2. Changing target consonant
To investigate whether the target consonant is crucial in the perception of words with
plain and emphatic consonants, the target consonants were swapped in this condition.
Subjects were presented with one of two types of words: a plain word with the
emphatic target consonant, or an emphatic word with the plain target consonant. If the
target consonant is the key, then perception of emphatic words is expected to follow the
emphatic consonant regardless of the vocalic environment. This condition was used for
fricatives and stops in the context of the low front vowel /a/ and the high back vowel
/u/. In condition 2 in Table 2 /T¿ab/ and /tAB/ the emphatic target consonant was placed
in the plain word and the plain target consonant was spliced in the emphatic word. In
/T¿ab/, 94.5% reported hearing /t/ and only 5.5% reported hearing /t¿/. In /tAB/, 66%
reported a /t¿/ and 34% reported a /t/. The same condition was applied to words with the
high back vowel, /u/. In /T¿ub/ 52.5% responded with a /t/ and 42.5% responded with a
/t¿/. 5% responded that they heard /s¿/. In /tUB/, 65.5% reported that they heard /t/ and
29.5% reported that they heard /t¿/. Similar to the case in /T¿ub/, 5% reported hearing
/s/.
Figure 1 shows these results. Letters in upper case represent the emphatic
segments.
126
100
94.5
% responses
80
66
60
65.5
52.5
%t
42.5
34
40
%t¿
29.5
20
5.5
0
T¿ab
tAB
T¿ub
tUB
Figure 1: percent /t/ and /t¿/ responses for words where target consonants were swapped
The same conditions in fricatives yielded similar results. For /S¿ab/, 93.5% reported
hearing an /s/ and only 1.5% said they heard an /s¿/. For /sAB/, 95% reported hearing
an /s¿/ and no one reported hearing an /s/. 2% reported hearing /t/ and 3% reported
hearing /t¿/. In the vowel /u/ environment, detection of plain and emphatic target
consonants is affected. In /S¿ub/, only 24% reported hearing /s/ and 76% reported
hearing /s¿/. In /sUB/, 53.5% reported hearing /s/ and 46.5% reported hearing /s¿/.
Figure 2 shows these results.
127
100
95
93.5
76
% responses
80
53.5
46.5
60
%s
%s¿
40
24
20
1.5
0
0
S¿ab
sAB
S¿ub
sUB
Figure 2: percent /s/ and /s¿/ responses for words where target consonants were swapped
In terms of reaction time, subjects responded significantly faster to detecting the plain
stops (1094 ms) than their emphatic counterparts (1379 ms) in the context of low front
vowels ([F (1, 398) = 15.796, p < 0.001]). With high back vowels, subjects detected
plain stops faster than their emphatic counterparts, (1255 ms) and (1299 ms),
respectively, but this difference was not significant, [F (1, 398) = 0.383, p = 0.536]. See
Figure 3.
128
1400
1379
1299
1300
1255
*
RT (ms)
1200
1094
1100
1000
900
800
T¿ab
tAB
T¿ub
tUB
Figure 3: Mean reaction time for /t/ and /t ¿/ when target consonants are swapped
For fricatives, similar results are obtained in the context of low front vowels but
opposite results are obtained in the context of high back vowels. Subjects detected /s¿/
with a mean RT of 986 ms and /s/ with a mean RT of 1059. This difference is
significant, [F (1, 398) = 15.796, p < 0.001]. This is not detected in the context of the
high back vowel. Subjects’ mean RT for detection of /s/ was 1098 ms and 1125 ms for
/s¿/. This difference is not significant, [F (1, 398) = 0.390, p = 0.533]. Figure 4 shows
these results.
129
1200
1125
1098
RT (ms)
1100
1000
1059
986
*
900
800
S¿ab
sAB
S¿ub
sUB
Figure 4: Mean reaction time for /s/ and /s ¿/ when target consonants are swapped
The confidence rating data show that subjects were largely confident of their responses.
Confidence ratings ranged from 4.5 for swapping the target stop to 4.9 for swapping the
low front vowel and the emphatic fricative. Table 4 shows these ratings for the different
conditions of changing the target consonant.
Table 4: Mean confidence for responses to changing the target consonant
Condition
tAB
T¿ab
tUB
T¿ub
sAB
S¿ab
sUB
S¿ub
total
Confidence rating
4.6
4.6
4.5
4.5
4.9
4.9
4.7
4.6
4.7
130
4.3.2.1. Summary
For /tab/ and /t¿ab/, when the emphatic target consonant was spliced in /tab/, 94.5%
subjects reported a plain /t/, and only 5.5% reported an emphatic /t¿/. However, in the
mirror image condition: the plain target consonant in /t¿ab/, only 66% subjects reported
an emphatic /t¿/ and 34% reported a plain /t/. When this manipulation is applied to /t/
and /t¿/ in /tub/ and /t¿ub/, that is, when the emphatic target consonant from /t¿ub/ was
spliced in /tub/, only 52.5% reported a plain /t/ and 42.5% reported /t¿/. In the mirror
image condition: /t/ in /t¿ub/, 65.5% reported a plain /t/ and 29.5% reported an emphatic
/t¿/. In fricatives, for /sab/ and /s¿ab/, when the emphatic target consonant was spliced
in /sab/, 93.5% subjects reported a plain /s/, and only 1.5% reported an emphatic /s¿/. In
the mirror image condition: the plain target consonant in /s¿ab/, 95% subjects reported
an emphatic /s¿/ and no one reported a plain /s/. When this manipulation is applied to /s/
and /s¿/ in /sub/ and /s¿ub/, that is, when the emphatic target consonant from /s¿ub/ was
spliced in the /sub/, only 24% reported a plain /s/ and 76% reported /s¿/. In the mirror
image condition: /s/ in /s¿ub/, 53.5% reported a plain /s/ and 46.5% reported an
emphatic /s¿/.
With regard to reaction time, subjects detected the plain stops faster than the emphatic
ones in the context of low front vowels. No significant differences were obtained in the
context of high back vowels. For fricatives, subjects detected the emphatic fricatives
faster than their plain counterparts in the context of low front vowels. No significant
differences were obtained in the context of high back vowels.
131
Finally, confidence ratings are still high under all of the manipulations in this condition.
A total of 4.7 confidence rating is scored for the condition of changing the target
consonant.
4.3.3. Changing target vowel
In this condition, the vowel in the same syllable with the target consonant, also referred
to as the target vowel, was cut and swapped. Therefore, the conditions resulting from
this manipulation are the plain word with the vowel from its emphatic counterpart and
the emphatic word with the vowel from the plain one. The perception of emphasis is
expected to follow the target vowel in these conditions – in line with the findings from
the acoustic study. While this finding is not expected to be affected by the manner of
the target consonant, the quality of the target vowel is expected to have an effect on the
perception of emphasis. It is expected that emphasis is perceived more clearly in the
presence of low front vowels than in the presence of high back vowels.
For / tAb /, only 6.5% responses were /t/ and an overwhelming 93.5% responses were
/t¿/. For / T¿aB /, subjects reported hearing 97.5% /t/ and only 2.5% /t¿/. For /tUb/, 69%
responses reported hearing a /t/ and only 26% reported hearing /t¿/. When the plain /u/
is spliced in the emphatic word, /T¿uB/, 53% reported hearing a /t/ and 41.5 reported
hearing a /t¿/. See Figure 5.
132
100
97.5
93.5
% responses
80
69
60
53
%t
41
%t¿
40
26
20
2.5
6.5
0
T¿aB
tAb
T¿uB
tUb
Figure 5: percent /t/ and /t¿/ responses for words where target vowels were swapped
The results obtained for fricatives are similar. In /sAb/, 91.5% responses were /s¿/ and
only 3% were /s/. In /S¿aB/, only 3% responses were /s¿/ and 92% were /s/. In the
environment of high back vowels, responses are split for /sUb/: 50% /s/ and 50 % /s¿/
but in /S¿uB/, 95.5% responses were /s/ and only 4.5% responses were /s¿/. Figure 6
shows these findings.
133
100
92
91.5
95.5
% responses
80
60
50 50
%s
%s¿
40
20
3
4.5
3
0
S¿aB
sAb
S¿uB
sUb
Figure 6: percent /s/ and /s¿/ responses for words where target vowels were swapped
The reaction times between these conditions did not show any significant differences.
Subjects detected /t¿/ in /T¿aB/ with a mean RT of 1083 ms and 1097 ms for /t/ in /tAb/.
Also, similar reaction times are obtained in the context of high back vowels: 1255 ms
for /t¿/ in /T¿uB/ and 1289 ms for /t/ in /tUb/. Figure 7 shows these results.
1400
1300
1255
1289
RT (ms)
1200
1100
1083
1097
T¿aB
tAb
1000
900
800
T¿uB
tUb
Figure 7: Mean reaction time for /t/ and /t ¿/ when target vowels are swapped
134
For stops, the reaction time data do not show significant differences. In the context of
low front vowels, mean reaction time for /s¿/ in /S¿aB/ was 1079 ms and 1058 ms for /s/
in /sAb/. In the context of high back vowels, subjects detected /s/ in /sUb/ faster than
they detected /s¿/ in /S¿uB/: 1259 and 1311 ms, respectively. See Figure 8.
1400
1311
1259
1300
RT (ms)
1200
1100
1079
1058
1000
900
800
S¿aB
sAb
S¿uB
sUb
Figure 8: Mean reaction time for /s/ and /s ¿/ when target vowels are swapped
The confidence ratings for the sound files in this condition were also very high. Similar
to the previous condition where the target consonant was spliced, subjects responded
with a high confidence level to sound files under this condition. Table 5 shows these
results.
135
Table 5: Mean confidence for responses to changing the target vowel
Condition
tAb
T¿aB
tUb
T¿uB
sAb
S¿aB
sUb
S¿uB
total
Confidence rating
4.9
4.7
4.5
4.6
4.8
4.9
4.6
4.7
4.7
4.3.3.1. Summary
In stops, when the plain vowel /a/ is spliced in the emphatic word /t¿ab/, 97.5% subjects
reported hearing a plain /t/, and only 2.5% reported the hearing the emphatic /t¿/. When
the emphatic vowel /a/ is spliced in the plain word /tab/, 93.5% reported an emphatic
/t¿/ and only 6.5% reported hearing /t/. The same conditions for /t¿ub/ and /tub/ yielded
different results. When the plain vowel /u/ is spliced in the emphatic word /t¿ub/, 53%
reported a plain /t/ and 41% reported an emphatic /t¿/. When the emphatic vowel /u/ is
spliced in the plain word /tub/, 69% reported a plain /t/ and only 26% reported an
emphatic /t¿/. In fricatives, when the plain vowel /a/ is spliced in the emphatic word
/s¿ab/, 92% subjects reported hearing a plain /s/, and only 3% reported the hearing the
emphatic /s¿/. When the emphatic vowel /a/ is spliced in the plain word /sab/, 91.5%
reported an emphatic /s¿/ and only 4.5% reported hearing /s/. The same conditions for
/s¿ub/ and /sub/ yielded similar results in one manipulation and different results in
another. When the plain vowel /u/ is spliced in the emphatic word /s¿ub/, 95.5%
reported a plain /s/ and 4.5% reported an emphatic /s¿/. However, when the emphatic
136
vowel /u/ is spliced in the plain word /sub/, subjects were equally split between /s/ and
/s¿/ reporting 50% each.
Reaction time data for this condition did not yield any significant differences. Subjects
reacted faster to emphatic stops than to their plain counterparts but slower to emphatic
fricatives than to their plain counterparts. None of these differences was significant.
The same pattern in terms of confidence rating is obtained here. The lowest confidence
ratings were obtained for swapping the target high back vowel in the word with the stop
consonant, 4.5. The highest ratings were obtained for swapping the target low front
vowel and the target emphatic fricative, 4.9.
4.4. Discussion
The results for the original target words show that they were detected at high accuracy
rates ranging from 99% to 100% expected responses. This finding suggests that splicing
the word into its segments and reassembling these segments does not affect the
detection rates considerably, which enables us to eventually conclude that different
responses to the other two conditions would not be due to splicing as opposed to
swapping the target consonants and vowels. Detection percentages for spliced words
did not seem to have been influenced by vowel quality. Subjects responded with similar
detection rates to words that had low front vowels and high back vowels. The
significant differences in reaction times, however, seem to be harder to explain.
137
Subjects detected plain stops faster than emphatic stops and plain fricatives slower than
emphatic fricatives – both in the context of low front vowels. This difference might not
be related to emphasis per se. It could be related to a possibility that /tab/ and /s¿ab/ are
more commonly frequent in the language than /t¿ab/ and /sab/, which seems to hold
water, from an impressionistic perspective, given the lack of word frequency charts for
UJA.
The difference in reaction times reported for words with low front vowels is absent in
high back contexts. This discrepancy in reaction times is compensated by the results on
the confidence ratings for these words. The confidence ratings obtained for this
condition ranged from a low of 4.7 to a high of 4.9, which indicates that subjects were
highly confident on their answers.
With regard to changing the target consonant, a few interesting findings can be
reported. When the emphatic stop is spliced into the plain word, a vast majority of the
responses, 94.5%, recognized a plain stop. The mirror image, though at a lesser
magnitude, is obtained when the plain stop is spliced in the emphatic word: 66%
reported hearing an emphatic stop. This suggests that the target consonant does not
carry the effects of emphasis in the environment of low front vowels. That is,
perception of emphasis does not depend on the properties of the emphatic stop. The
same results are fostered by the findings on the fricatives. In conditions identical to the
ones mentioned here, subjects reported hearing a plain fricative when an emphatic
138
fricative was spliced in a plain word, and reported hearing an emphatic fricative when a
plain one was spliced in an emphatic word. Therefore, we can state that regardless of
the manner of articulation for the target consonant, the perception of emphasis is not
driven by the consonant in the environment of low front vowels.
Contrary to the findings on manner of articulation with regard to emphasis perception,
vowel quality seems to be critical in influencing the perception of emphasis. As
mentioned in the acoustic study, see Chapter 2, the mean difference in the second
formant frequency between low front vowels in plain and emphatic environments is
larger than that in high back vowels. This difference maps on to perception. Based on
the frequency of target consonant detections, subjects found it harder to decide whether
they heard a plain or an emphatic /t/ when these two stops were swapped in / T¿ub/ and
/Tub/. However, when the plain stop was spliced into the emphatic word, 65.5%
responses reported a plain /t/ compared 29.5% /t ¿/ responses. On the other hand,
subjects showed preference for /s¿/ (76%) when the emphatic fricative was spliced in
the plain word in the environment of high back vowels. These results show that the
high back vowels tend to impede the clear effects of emphasis because the magnitude
of F2 lowering in emphatic environments is smaller compared to that in low front
vowels. The findings on the reaction time data support the overriding influences of low
front vowels compared to the lesser effects of high back vowels. In stops and fricatives,
differences in reaction times to plain and emphatic segments were significant only in
the environment of low front vowels. They were, however, contradictory. While
subjects detected plain stops faster than emphatic stops, they detected emphatic
139
fricatives faster than plain fricatives. No significant differences were found in the
context of high back vowels. Despite these findings on reaction time, subjects thought
they were confident of their answers. The lowest end on the confidence ratings scale
was at 4.5 and the highest was at 4.9. Both of these ratings are substantially indicative
of high confidence.
In the third condition, swapping the target vowels, the perception results support the
overriding effects of the vowels in emphasis. When the vowel from the plain word is
spliced into the emphatic word, subjects detected the plain stop /t/ with a very high
percentage. When the vowel from the emphatic word was spliced into the plain word,
subjects detected the emphatic stop /t¿/. These results were also true for the fricatives.
In both cases for stops and fricatives, these results are obtained clearly in the context of
low front vowels. In the context of high back vowels, subjects’ responses were close to
chance level. However, the results for the vowels show some differences from those
obtained for the stops. When the plain vowel was spliced into the emphatic word,
95.5% responses reported a plain /s/, meaning that the high back vowel /u/ did
influence subjects’ perception. The mirror image of this manipulation, splicing the
emphatic vowel in the plain word, yielded perfect chance level responses split between
the plain and emphatic fricative detections. In terms of reaction time data, no significant
differences have been found under the different manipulations in this condition. One
finding is that reaction times were generally shorter for words that had the low front
vowel compared to those that had the high back vowel. As for the confidence ratings,
140
the results follow the pattern in the first two conditions. Subjects respond with high
confidence ratings averaging 4.7.
While the results are neat and tidy in the environment of the low front vowel /a/, the
change in vowel quality to /u/ does not yield consistent results. In the environment of
the vowel /a/, wherever the vowel is spliced, subjects’ perception is driven by it. So, if
the vowel comes from an emphatic word, the perception of the target consonant is
largely /t¿/ or /s¿/, and vice versa. In the context of the high back vowel /u/, these results
do not form a pattern. Sometimes subjects report hearing a plain consonant by high
percentages and sometimes their perception is close to or at a chance level. The
intrinsic characteristics of the high back vowel, namely that the difference between F1
and F2 is too small compared to that in /a/, leaves less room for F2 to decline and F1 to
rise in the emphatic environments. This leads to inconsistent results in the perception of
emphasis under the conditions presented in this chapter. Another consistent result that
supports this finding is that reaction times to words with low front vowel were
consistently shorter compared to these with high back vowels.
The results of this study show that perception of emphasis is not driven by the target
consonant. Although emphasis is triggered by the existence of an emphatic segment,
which is always an obstruent, its effects are observed on the vowel. The low front
vowel is more susceptible to emphasis than the high back vowel. While emphasis is
clearly realized on the low front vowel, it is not as salient in the context of high back
141
vowels. This is due to the fact that there is more room for F1 rising and F2 lowering in
the context of low front vowels than in the context of high back vowels.
142
Chapter Five
Summary, discussion and conclusion
The present study examined the acoustic and perceptual correlates of emphasis in
Urban Jordanian Arabic. The study consisted of three main experiments: two acoustic
experiments: one on monosyllabic words, and another on bisyllabic words, and a
perception experiment. In the acoustic experiment on monosyllables, four pairs of plain
and emphatic consonants were used along with the six vowels of UJA to create a list of
48 minimal pairs of monosyllabic words of the CVC structure. The target consonants
were either word-initial or word-final. The duration of the plain and emphatic
consonants was measured, and their spectral moments and locus equations were
calculated. The three formant frequencies of the vowel, F1, F2, and F3 were measured
at onset, temporal midpoint and offset of the vowel. As for the acoustic experiment on
bisyllables, the wordlist consisted of 80 target CV.CVC pairs where the target
consonant was word-initial, word-medial or word-final. The perception of emphasis
was also investigated by means of splicing and swapping the target consonants and
vowels. Four pairs of words were selected from the stimuli used in the acoustic
experiment. The perception experiment was used to test whether the results in the
acoustic study are perceptually valid.
143
5.1. Summary of results
5.1.1. Acoustic study: the monosyllables
The results for the acoustic study show that emphatic consonants are longer than their
plain counterparts. This difference is, however, limited to fricatives in word-final
positions. Word-initial and word-final stops do not show significant differences in plain
and emphatic positions. Additionally, fricatives in word-initial positions are not
significantly different. Target vowel duration was not significantly different in plain
and emphatic environments. Only one of the four spectral moments measured showed
significant differences; the center of gravity is significantly higher for emphatic stops
and lower for emphatic fricatives in word-final positions. Locus equations showed that
emphatic consonants had a lower mean slope compared to their plain counterparts
while y-intercept values showed no significant differences.
As expected, vowel F1 was significantly higher in emphatic environments. However,
this effect was strongest closest to the target consonant. There was often a tendency for
this effect to fade as the measurement was taken further away from the target
consonant. F2 was consistently lower in emphatic environments. The results for F3
were similar to those attested for F1: F3 increased in emphatic environments but this
increase faded away as the measurement was taken further away from the target
consonant.
Several
interactions
were
observed
for
the
formant
frequency
measurements. The formant frequency values were often affected by vowel quality.
High front and low front vowels were less resistant to emphasis effects compared to
144
high back vowels. In other words, where the formant frequency was expected to
increase or decrease, it did so by a greater extent in the environment of high and low
front vowels and by a lesser extent in the environment of high back vowels. Sporadic
interactions were found for position of the target consonant, manner of articulation,
vowel length and word type. No interactions were observed for voicing and gender.
5.1.2. Acoustic study: the bisyllables
Measurements on consonant and vowel durations and spectral moments for consonants
did not reveal consistent differences. Measurements on the bisyllables focused on
spectral measurements for vowels. For most of the data, vowel F1 and F3 significantly
increase but vowel F2 significantly decreases in emphatic environments. Formant
values usually show these results when the measurements are taken for the target
vowels: that is, vowels in the same syllable with the target consonants. As the
measurements are taken outside of the target vowel, effects of emphasis tend to fade
away. This is clearer when the measurements are taken on the left rather than right of
the target vowel, thus showing more prominence for anticipatory emphasis spread.
Furthermore, the effects of emphasis are clearer in the environments of high and low
front vowels compared to high back vowels. This was also attested for monosyllables.
Finally, measurements on vowel formant frequency show greater plain – emphatic
differences for males than females, thus suggesting that emphasis is more salient in
male speech.
145
5.1.3. The perception study
As for the perception study, the two types of consonants, stop and fricative, were
included along with two types of vowel quality: low front and high back. Three
conditions were used in this study: the original words spliced, the target consonant
swapped, and the target vowel swapped. Results for the first condition show that
subjects responded with a high degree of accuracy. That is, they identified each target
consonant as expected with an accuracy rate ranging from 99 – 100%. In the second
condition, results showed that subjects identified the consonant based on the vocalic
environment and not on the target consonant itself. If the target consonant was plain but
spliced in an emphatic vocalic environment, subjects reported an emphatic consonant,
and vice versa. The mirror image was detected for the third condition – splicing the
target vowel. If the vowel came from the emphatic word and was spliced into the plain
environment, subjects detected an emphatic target consonant, and vice versa. These
results, however, were influenced by vowel quality. If the vowel was low front, it
carried the effects of emphasis. If the vowel was high back, in most cases, responses
were close to a chance level.
5.2. Discussion
The results of the present study are in line with previous research on emphasis in the
areas where similar measurements and procedures have been adopted. Previous
findings for the duration of the target consonant showed that emphatic consonants were
146
significantly longer than plain ones, see Obrecht (1968). In the present study, the target
vowel durations were not significantly different. There have been two accounts in the
literature with contradicting findings for target vowel durations. The acoustic study of
the bisyllables shows that durational differences are not consistently influenced by the
presence of emphatic consonants. It seems that consonant and vowel duration are not a
stable cue for emphasis, especially in the case of short vowels. The allophonic changes
in the Arabic vowels seem to be easier to capture in quality rather than quantity. The
investigations regarding the spectral moments did not yield significant differences.
Apart from the findings on the center of gravity, which showed significant differences
in monosyllables, none of the other moments turned out to show significant differences.
While Norlin (1987) found significant lowering in the center of gravity for word-initial
emphatic sibilants, the present study reports such findings in the same direction only for
word-final fricatives and in the opposite direction for word-final stops. None of the
moments was significantly different in word-initial position. It should be noted here
that the measurements for spectral moments were characterized by a wide degree of
variability depending on the point at which the moments were calculated. This
variability was inevitable due to the differences in burst duration for stops. Based on
findings for the monosyllables, and due to the unclear effect of consonantal spectral
moments, moments were not reported in the study on the bisyllables.
As reported by almost all studies that investigated the acoustics of emphasis, vowel F2
turns out to be a very salient cue for emphasis. Vowel F2 drops in emphatic positions
147
compared to their plain counterparts. This is true for vowels in monosyllables and in
bisyllables as well. In monosyllables, Vowel F2 in words with initial target consonants
drops by 516 Hz, 253 Hz, and 165 Hz at onset, temporal midpoint and offset positions,
respectively. For words with final target consonants, F2 drops by 228 Hz, 248 Hz and
462 Hz at onset, temporal midpoint and offset positions, respectively. It should be
noted here that we are looking at the mirror image in word-initial and word-final
positions. While F2 onset is closest to the target consonant in words with initial
emphatics, F2 offset is closest to the target consonant in words with final emphatics.
This points to the gradient nature of emphasis spread within the same word, (cf.
Zawaydeh 1999; Al-Masri and Jongman 2004). F1 and F3 are higher in emphatic
positions compared to their plain counterparts. While the gradient effect is evident in
the degree of drop in F2 in emphatic positions, it is similarly evident in the results for
F1 and F3 in that the increase in emphatic positions diminishes as the measurement is
taken further away from the target consonant. Additionally, the extent of emphasis was
not similar for all vowels. It is worth noting that emphasis is realized more in the
environment of high front and low front vowels compared to high back vowels where
the relatively compact spectrum for /u/ limits the degree of formant increase or
decrease. This leads us to predict that high back vowels are more likely to block
emphasis spread compared to high front and low front vowels.
148
In terms of locus equations, the set of emphatic consonants in UJA patterned like it did
in other dialects of Arabic (Yeou, 1997) with flatter slopes compared to their plain
counterparts.
In bisyllables, the results pattern similarly. In addition, measurements on the bisyllables
give us a clearer idea about emphasis spread. It shows that emphasis clearly crosses the
left syllable boundary but is confined in the right direction to the target syllable.
Further, measurements on the bisyllables show a clear influence of vowel quality,
whereby the high back vowel /u/ seems to be a stronger blocker compared to the high
front and high back vowels. Finally, the magnitude of the plain-emphatic difference
suggests a greater tendency among males for pronouncing emphasis compared to
females.
The third experiment illustrates the perceptual consequences of the findings in the
acoustic study. The most salient findings on the acoustics of emphasis carried over to
perception, indicating that the target vowel drives the perception of emphasis. In
addition to this result, the more intricate findings on vowel quality materialized in
perception. Perception of emphasis was carried by low front vowels but not by high
back vowels. Despite the differences in the experimental design, these findings
replicate reports in the literature (Obrecht, 1968).
5.3. Conclusion
The findings reported in this study are significant in several ways. Experimental
research on Arabic has been limited compared to other languages. It is only recently
149
that more similar studies are reported. The previous literature on Arabic, especially the
literature written in Arabic, has been mostly descriptive. The present study lends itself
to the experimental domain. The findings reported in the present study should be
incorporated in the pedagogical fields, especially now that textbooks on Arabic are in
demand. In the absence of the emphasis distinction in most world languages, it is
crucial that the methods adopted in presenting emphasis as a phonemic distinctive
feature in Arabic be relevant to the tools available for the second language learner.
Rather than focusing on the differences between the consonants, which is how most
native speakers of Arabic think they perceive emphasis, the differences in the vowels
should be highlighted.
5.4. Follow-up
This study is basically one step on a long way. Several aspects of emphasis are still
wanting. While the acoustic correlates of monosyllables and bisyllables have been
systematically presented, analysis of emphasis spread beyond the neighboring syllable
is still lacking. Studying the acoustics of trisyllables should complete the picture. It
should allow us to understand the extent and directionality of emphasis spread, possible
consonantal and vocalic segments that block emphasis spread or limit it to its minimum
domain. Moreover, the investigation of the possible role of amplitude and spectral
moments and locus equations for consonants in adjacent syllables is warranted.
Studying the effects of these factors on the perception of emphasis is yet another
possible area of investigation. In addition to the investigation of emphasis in
150
trisyllables, geminate and adjacent emphatics have not been discussed in the literature
so far. A follow-up study using these data would possibly complete the picture.
This is the first step in a series of more acoustic and perceptual experiments to cover all
the unexplored characteristics of the emphatic signal. Conducting these experiments
should allow for much needed contributions to the field of Arabic linguistics.
151
References
Al-Ani, S. (1970) Arabic Phonology. The Hague: Mouton.
Ali, L. & Daniloff, R. (1974) A contrastive cinefluorographic investigation of the
articulation of the emphatic – nonemphatic cognate consonants, Studia Linguistica, 26,
81 – 105.
Alioua, A. (1995). L'effet des consonnes d'arriere et des emphatiques sur la nature
acoustique des voyelles longues de l'Arabe litteral moderne. Unpublished doctoral
dissertation, Universite Laval, Montreal, Canada.
Al-Masri, M. and Jongman, A. (2004) Acoustic correlates of emphasis in Jordanian
Arabic: preliminary results. In Augustine Agwuele et al, (ed.)
Proceedings of the
2003 Texas Linguistics Society Conference. Somerville, MA: Cascadilla Proceedings
Project, 96-106.
Alosh, M. (1987) The perception and acquisition of pharyngealized fricatives by
American learners of Arabic and implications for teaching Arabic phonology. PhD
Dissertation, Ohio State University.
Alwan, A. (1983) Acoustic and perceptual correlates of pharyngeal and uvular
consonants. PhD Dissertation, MIT.
Alwan, A. (1989) Perceptual cues for place of articulation for the voiced pharyngeal
and uvular consonants, Journal of the acoustic society of America,
86, 549 – 556.
Bani Yasin, R and Owens, J. (1987) The phonology of a Northern Jordanian Arabic
dialect, ZDMG, 137, 291 – 331.
152
Card, E. (1983) A Phonetic and Phonological Study of Arabic Emphatics. PhD
Dissertation, Cornell University, University Microfilms International, Ann Arbor:
Michigan.
Davis, S. (1995) Emphasis in Grounded Phonology, Linguistic Inquiry, 26, 465 – 498.
Chiba, T. & Kajiyama, M. (1942) The vowel: Its nature and structure. Phonetic Society
of Japan: Tokyo
Delattre, P. (1971) Pharyngeal Features in the Consonants of Arabic, German, Spanish,
French and American English, Phonetica. 23, 129 – 155.
El-Halees, Y. (1985) The role of F1 in the place-of-articulation distinction in Arabic,
Journal of Phonetics, 13, 287 – 298.
El-Halees, Y. (1987) F2 and Back Articulation in Arabic, Journal of the College of
Arts, King Saud University, 14, 21 – 44.
Ferguson, C. (1956) The Emphatic l in Arabic, Language. 32, 446 – 452.
Ghowail, Thanaa Ibrahim (1987) The acoustic phonetic study of two pharyngeals /š, ¿/
and two laryngeals /½,h/ in Arabic. PhD Dissertation, Indiana University.
Giannini, A. and Pettorino, M. (1982) The emphatic consonants in Arabic. Istituto
Universitario Orientale: Napoli.
Herzallah R. (1990) Aspects of Palestinian Arabic Phonology: A non-linear approach.
PhD Dissertation, Cornell University.
Jongman, A., Wayland, R., & Wong, S. (2000) Acoustic characteristics of English
fricatives, Journal of the Acoustical Society of America. 108, 1252 – 1263.
153
Kahn, M. (1975) Arabic Emphasis: The Evidence for Cultural Determinants of
Phonetic Sex-Typing, Phonetica, 31, 38 – 50.
Kent, R. and Read, Ch. (1992) The Acoustic Analysis of Speech. Singular Publishing
Group: California.
Khattab, G., Al-Tamimi, F. & Heselwood, B. (2006) Acoustic and Auditory differences
in the /t/-/t/ opposition in male and female speakers of Jordanian Arabic. In Boudelaa,
S. (ed.) Perspectives on Arabic Linguistics XVI. John Benjamins Publishing Company:
Amsterdam/Philadelphia
Ladefoged, P., & Maddieson, I. 1996. The Sounds of the World’s Languages. Oxford:
Blackwell
Laufer, A. and Baer, T. (1988) The emphatic and pharyngeal sounds in Hebrew and in
Arabic. Language and Speech. 31, 181 – 205.
Laver, John. 1994. Principles of phonetics. Cambridge: Cambridge University Press.
Lehn, W. (1963) Emphasis in Cairo Arabic, Language 39, 129 – 139.
McCarthy, J. (1994) The phonetics and phonology of Semitic pharyngeals. In Norlin,
K. (1985) Pharyngealization in Cairo Arabic, Working papers, 28, Lund University:
Sweden.
Norlin, K. (1987) A phonetic study of emphasis and vowels in Egyptian Arabic. PhD
dissertation. Lund University: Sweden
Obrecht, D. (1968) Effects of the 2nd formant on the perception of velarized consonants
in Arabic. The Hague, Mouton, Paris.
154
Obrecht, D (1961) Effects of the second formant on the perception of velarized
consonants in Arabic. PhD Dissertation, University of Pennsylvania.
Royal, A. (1985) Male/female pharyngealization patterns in Cairo Arabic: a
sociolinguistic study of two neighborhoods. PhD Dissertation, University of Texas at
Austin.
Semaan, K. (1968) Linguistics in the Middle Ages: Phonetic Studies in Early Islam.
Leiden: E. J. Brill.
Suleiman, S. (1985) Jordanian Arabic between Diglossia and Bilingualism: Linguistic
analysis. John Benjamins Publishing Company: Amsterdam/Philadelphia
Sussman, H. et al. (1991) An investigation of locus equations as a source of relational
invariance for stop place categorization, Journal of the Acoustical Society of America,
90, 1309 – 1325.
Versteegh, K. (2001) The Arabic Language. Edinburgh University Press.
Wahba, K. (1993) A sociolinguistic treatment of the feature of emphasis in Egypt. PhD
Dissertation, Alexandria University: Egypt.
Watson, J. (1999) Remarks and Replies: the Directionality of Emphasis in Arabic.
Linguistic Inquiry, 30, 289 – 300.
Watson, J. (2002) The Phonology and Morphology of Arabic. Oxford UP.
Yeou, M. (1995) Trading relations between cues for the pharyngealized/non
pharyngealized contrast, Proceedings of the 13th ICPhSc, KTH and Stockholm
University, Stockholm, 464 – 467.
155
Yeou, M. (1997) Locus equations and the degree of coarticulation of Arabic
consonants. Phonetica, 54, 187 – 202.
Younes, M. (1994) On emphasis and /r/ in Arabic. In Hole, C. and Eid, M. (eds.)
Perspectives on Arabic linguistics, Vol. VI. Amsterdam: John Benjamins.
Zawaydeh, B. (1999) The Phonetics and Phonology of Gutturals in Arabic. PhD
Dissertation, Indiana University.
Zemánek, P. (1996) The origins of Pharyngealization in Semitic. Praha: Enigma
corporation, Ltd.
Zuraiq, W., & Zhang, J. (2006). Phonological assimilation in Urban Jordanian Arabic.
Kansas Working Papers in Linguistics, 28, 33-64.
156
Appendices
Appendix I: The acoustic study: the monosyllables:
Monosyllables: target word-initial
No
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Plain
sab
sib
sub
sa:b
si:b
su:b
tab
tib
tub
ta:b
ti:b
tu:b
dab
dib
dub
da:b
di:b
du:b
ðab
ðub
ðib
ða:b
ðu:b
ði:b
Gloss
He cursed
He cursed
You curse
He left
You leave
Nonword
He perished
Nonword
You repent
He repented
Nonword
You repent
He fell
Nonword
Bear
It melted
Bear
You melt
Nonword
You protect
Nonword
It melted
You melt
Wolf
Emphatic
s¿ab
s¿ib
s¿ub
s¿a:b
s¿i:b
s¿u:b
t¿ab
t¿ib
t¿ub
t¿a:b
t¿i:b
t¿u:b
d¿ab
d¿ib
d¿ub
d¿a:b
d¿i:b
d¿u:b
ð¿ab
ð¿ub
ð¿ib
ð¿a:b
ð¿u:b
ð¿i:b
157
Gloss
He poured
Nonword
You pour
He scored
You score
Nonword
It touched
You recover
Nonword
He recovered
You recover
Brick
Nonword
Nonword
You keep
Nonword
Nonword
Nonword
Nonword
Nonword
Nonword
Nonword
Nonword
Nonword
Structure
CVC
Monosyllables: target word-final
No
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Plain
bas
bis
bus
ba:s
bi:s
bu:s
bat
bit
but
ba:t
bi:t
bu:t
bad
bid
bud
ba:d
bi:d
bu:d
bað
bið
buð
ba:ð
bi:ð
bu:ð
Gloss
Enough!
Cat
Nonword
He kissed
Nonword
You kiss
He decided
You decide
You decide
He slept
You sleep
Shoes
Big stone
Nonword
Nonword
It ended
You kill; end
Nonword
Nonword
Nonword
Nonword
Nonword
Nonword
Nonword
Emphatic
bas¿
bis¿
bus¿
ba:s¿
bi:s¿
bu:s¿
bat¿
bit¿
but¿
ba:t¿
bi:t¿
bu:t¿
bad¿
bid¿
bud¿
ba:d¿
bi:d¿
bu:d¿
bað¿
bið¿
buð¿
ba:ð¿
bi:ð¿
bu:ð¿
158
Gloss
He saw
Nonword
You see
Bus
Nonword
Nonword
Nonword
Nonword
Nonword
Armpit
shoes
Nonword
Nonword
Nonword
Nonword
It laid eggs
White (pl)
Nonword
Nonword
Nonword
Nonword
Nonword
Nonword
Nonword
Structure
CVC
Appendix II: The acoustic study: the bisyllables:
Rightward spread
Bisyllables: (target C word-initial)
No
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Plain
sabat
sibat
subat
saabat
siibat
suubat
thabat
thibat
thubat
thaabat
thiibat
thuubat
tabat
tibat
tubat
taabat
tiibat
tuubat
dabat
dibat
dubat
daabat
diibat
duubat
Gloss
he selpt
nonword
nonword
she left you
nonword
nonword
nonword
nonword
nonword
it melted
nonword
nonword
nonword
nonword
nonword
she repented
nonword
nonword
nonword
nonword
nonword
it melted
nonword
nonword
Emphatic
sabat
sibat
subat
saabat
siibat
suubat
thabat
thibat
thubat
thaabat
thiibat
thuubat
tabat
tibat
tubat
taabat
tiibat
tuubat
dabat
dibat
dubat
daabat
diibat
duubat
159
Gloss
nonword
nonword
nonword
she touched you
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
she recovered
kindness
his brick
nonword
nonword
nonword
nonword
nonword
nonword
Structure
CV.CVC
Leftward spread
Bisyllables: (target C word-final)
No
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Plain
tabas
tabis
tabus
tabaas
tabiis
tabuus
tabath
tabith
tabuth
tabaath
tabiith
tabuuth
tabat
tabit
tabut
tabaat
tabiit
tabuut
tabad
tabid
tabud
tabaad
tabiid
tabuud
Gloss
nonword
nonword
nonword
nonword
nonword
she kisses
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
she sleeps
she sleeps
nonword
nonword
nonword
nonword
nonword
nonword
nonword
Emphatic
tabas
tabis
tabus
tabaas
tabiis
tabuus
tabath
tabith
tabuth
tabaath
tabiith
tabuuth
tabat
tabit
tabut
tabaat
tabiit
tabuut
tabad
tabid
tabud
tabaad
tabiid
tabuud
160
Gloss
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
it lays eggs
nonword
Structure
CV.CVC
Bidirectional spread
Bisyllables: (target C word-medial)
No
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Plain
kasak
kasik
kasaak
kasiik
kisak
kisik
kisaak
kisiik
kathak
kathik
kathaak
kathiik
kithak
kithik
kithaak
kithiik
katak
katik
kataak
katiik
kitak
kitik
kitaak
kitiik
kadak
kadik
kadaak
kadiik
kidak
kidik
kidaak
kidiik
Gloss
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
Emphatic
kasak
kasik
kasaak
kasiik
kisak
kisik
kisaak
kisiik
kathak
kathik
kathaak
kathiik
kithak
kithik
kithaak
kithiik
katak
katik
kataak
katiik
kitak
kitik
kitaak
kitiik
kadak
kadik
kadaak
kadiik
kidak
kidik
kidaak
kidiik
161
Gloss
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
nonword
Structure
CV.CVC